Squash commits of GitHub pages

[fpoli]

This PR adds commands that should squash the deploy commits of the GitHub pages. If approved, I'll try executing these commands manually before merging.

These commits pollute the tree and are not particularly informative:

> git log --graph --abbrev-commit --decorate --format=format:'%h - (%ar) - %an%d' --all
...
| * 69195ede37c - (39 minutes ago) deploy: 0ed9c6248c41f5b20a7fc98c038fcbd1f961a157 - fpoli (origin/gh-pages)
| * 4836e1040a1 - (74 minutes ago) deploy: 0ed9c6248c41f5b20a7fc98c038fcbd1f961a157 - fpoli
| * b17270e6025 - (2 days ago) deploy: a2e70818ed3cccff2349fa2d4e0b199b0a25e8e4 - Aurel300 (gh-pages)
| * 244f8291338 - (2 days ago) deploy: a2e70818ed3cccff2349fa2d4e0b199b0a25e8e4 - Aurel300
| * 1d630a8ded5 - (3 days ago) deploy: fed995e5bc2c9e63efcebc5f3bb67b7515ff7a77 - bors[bot]
| * 45d151aa7aa - (3 days ago) deploy: fed995e5bc2c9e63efcebc5f3bb67b7515ff7a77 - bors[bot]
| * c454ae20e71 - (8 days ago) deploy: 34981398a0bd12b412ce800a008aa527172013be - Aurel300
| * 02452ddaa50 - (8 days ago) deploy: 34981398a0bd12b412ce800a008aa527172013be - Aurel300
| * 794a9dd3755 - (12 days ago) deploy: 2bf067333eae186e5300bc06416024369ef44fd4 - fpoli
| * df6a295eec4 - (12 days ago) deploy: 2bf067333eae186e5300bc06416024369ef44fd4 - fpoli
| * 6a622a0c6ef - (12 days ago) deploy: ba9b3ad256f1914cbc75e60fa8a8942dc202cb23 - fpoli
| * 3438989be45 - (12 days ago) deploy: ba9b3ad256f1914cbc75e60fa8a8942dc202cb23 - fpoli
| * c322b2ea7ab - (2 weeks ago) deploy: 701759f80d1aad47bbbf29b3e88c48eb4b20fa1f - fpoli
| * 93e3017647a - (2 weeks ago) deploy: 701759f80d1aad47bbbf29b3e88c48eb4b20fa1f - fpoli
| * 225e2e09518 - (2 weeks ago) deploy: ee91e7772e764c910b1e6638f609ad5da0a791a7 - fpoli
| * 027d3028ce2 - (2 weeks ago) deploy: ee91e7772e764c910b1e6638f609ad5da0a791a7 - fpoli
| * 2b29c3c2ba5 - (2 weeks ago) deploy: ca118f6e8c76570f476b2b88b8d7527c12ba3241 - Aurel300
| * 3b21ebf4c2e - (2 weeks ago) deploy: ca118f6e8c76570f476b2b88b8d7527c12ba3241 - Aurel300
| * 1eb9235be4d - (2 weeks ago) deploy: e8c81a6a2ba6674fa9cae81e7339abbac8e8b237 - bors[bot]
| * ff1f2199a9a - (2 weeks ago) deploy: e8c81a6a2ba6674fa9cae81e7339abbac8e8b237 - bors[bot]
| * 71225f4ce7d - (3 weeks ago) deploy: 734426eb50e6de7454b40694d5a8e843c05d9ee8 - fpoli
| * b3024f688c4 - (3 weeks ago) deploy: 734426eb50e6de7454b40694d5a8e843c05d9ee8 - fpoli
| * 6ef919ef42c - (3 weeks ago) deploy: 734426eb50e6de7454b40694d5a8e843c05d9ee8 - fpoli
| * 699e80bff24 - (3 weeks ago) deploy: 2f46f65f13d48c4fdf574cae33b7cf25f47785c2 - Aurel300
| * b05589b9a42 - (3 weeks ago) deploy: 2f46f65f13d48c4fdf574cae33b7cf25f47785c2 - Aurel300
| * 172b78c9893 - (3 weeks ago) deploy: 638ebd1d7eb755aeff571d50fda4fd948ddfe4e2 - fpoli
| * bccd8fe4e83 - (3 weeks ago) deploy: 638ebd1d7eb755aeff571d50fda4fd948ddfe4e2 - fpoli
| * e75522276f9 - (3 weeks ago) deploy: 693783f61ed2abe34788800b61ed2e84ac1e09d3 - fpoli
| * 53ea0a809fc - (3 weeks ago) deploy: 693783f61ed2abe34788800b61ed2e84ac1e09d3 - fpoli
| * 35bbec0554d - (3 weeks ago) deploy: 98d79c492aae33f57e72716fde4da69956abf5ab - fpoli
| * b09b900fe03 - (3 weeks ago) deploy: 503fe9bd598488ab15e08360c8b2637958507c0c - fpoli
| * 9e89502bd79 - (3 weeks ago) deploy: 98d79c492aae33f57e72716fde4da69956abf5ab - fpoli
| * e7cf926bbd4 - (3 weeks ago) deploy: 503fe9bd598488ab15e08360c8b2637958507c0c - fpoli
| * 7947d932bc3 - (4 weeks ago) deploy: 46488be654aa770797378929a9d1fd265b49aa29 - bors[bot]
| * 7b9831dcff2 - (4 weeks ago) deploy: 46488be654aa770797378929a9d1fd265b49aa29 - bors[bot]
| * aed504cf13d - (4 weeks ago) deploy: 3a6860cdbac39f83b07fe190b0077adcd4500b5c - fpoli
| * 793325de9b7 - (4 weeks ago) deploy: 3a6860cdbac39f83b07fe190b0077adcd4500b5c - fpoli
| * d01c73cd65d - (4 weeks ago) deploy: db0c9740acedf2b2b3d3710ab0937ba7754fa579 - fpoli
| * e1923ce6f0d - (4 weeks ago) deploy: f68d05a88b078ee73f3f2e3ae1ded170699adba2 - fpoli
...

[Aurel300]

@fpoli Can you see if force-orphan on the GH publish action does the same thing?

[fpoli]

I didn't notice force_orphan, thanks. However, it seems that it doesn't work with keep_files: peaceiris/actions-gh-pages#455.

[Aurel300]

Ok, three things:

• Unless I'm forgetting something, keep_files is used only for publishing the benchmark files to GH pages, which as of yet we are not displaying anywhere anyway. If we don't care about these (I don't think GH pages are the right place for this anyway), we can not use the keep_files option.
• Speaking of benchmarks, there is a bunch of them in that branch. Do we want to keep them? @vakaras ? Should we replace this with the perfbot @zgrannan ?
• (Very minor.) It seems the logo.png file got corrupted at some point? At least it does not display for me.

[fpoli]

• Unless I'm forgetting something, keep_files is used only for publishing the benchmark files to GH pages, which as of yet we are not displaying anywhere anyway. If we don't care about these (I don't think GH pages are the right place for this anyway), we can not use the keep_files option.

Good point. If we remove keep_files: false then we should make sure that destination_dir is set, or each of the workflows would delete the files published by the others.

• (Very minor.) It seems the logo.png file got corrupted at some point? At least it does not display for me.

Mmm, same here.

[fpoli]

• Unless I'm forgetting something, keep_files is used only for publishing the benchmark files to GH pages, which as of yet we are not displaying anywhere anyway. If we don't care about these (I don't think GH pages are the right place for this anyway), we can not use the keep_files option.

Good point. If we remove keep_files: false then we should make sure that destination_dir is set, or each of the workflows would delete the files published by the others.

Actually, docs.yml cannot use destination_dir because it needs to publish index.html in the root folder. I think our problem is that we have three workflows publishing files instead of just one. If we remove the benchmarks then we can merge the remaining two workflows.

[fpoli]

• Speaking of benchmarks, there is a bunch of them in that branch. Do we want to keep them?

@vakaras? If we remove them then I can merge and simplify the workflows.

[fpoli]

Instead of merging this PR, it seems from the discussion above that the best approach is to

1. Remove the benchmark workflow (which we never really used)
2. Merge the two remaining workflows, publishing with the flag force_orphan=true.