build(docs): include full changelog and all contributors

[agilgur5]

Partial fix for #13206 (comment)

Motivation

• there's a filter on the commits in the changelog generation script, but the CHANGELOG does not mention that

Modifications

• say "Selected Changes" to explicitly note that it is not full

  • and put a link to the Full Changelog/Full Diff via GH compare view above that

• include all Contributors to the release, not just the filtered ones

  • also collapse the list of contributors by default with the HTML <details> tag

Verification

I ran the changelog.sh script locally and committed the resulting CHANGELOG.md in this PR.

• See the rendered commit
• I also double-checked the entirety of the diff myself

[terrytangyuan]

Hmm I wonder why this shows up in addition to "Yuan Tang" below

[agilgur5]

It's probably because you had a different name in your commit author previously (i.e. Author: Yuan (Terry) Tang <email>. Since the script uses pure git (no GitHub so it can't do, for instance, email -> GH user), it pulls entirely from the existing commit metadata in the repo

[terrytangyuan]

Is there a way to hack/merge these different ones?

[agilgur5]

Not without making the script even hackier than it already is 😬
I would prefer to use GH IDs where possible, but that would require some non-trivial work to do

You'd effectively have to make a dictionary of author names to reference and choose one by some preference (I guess latest one), which means an associative array in Bash, which requires Bash 4+ (i.e. breaks on older Bash, older Macs, etc b/c Bash portability kinda sucks 😐 ). So another loop and a dictionary lookup.

I did notice this in the diff, but thought it was nbd, especially since this PR collapses the Contributors section anyway. Your own name here is though, so a bit different I get it

[agilgur5]

You'd effectively have to make a dictionary of author names to reference and choose one by some preference (I guess latest one), which means an associative array in Bash

Oh wait, the author name and email are the (non-unique) identifier, so you'd actually need an associative array for old name + email to new name + email. There could be multiple matches for that and you might not be able to match if both the email and name are different -- which they very well might be in your case if you were using corporate emails.
In other words, quite hacky 🙃

[agilgur5]

IMO, I'd punt this until we can do GH ID lookups, but that's not perfect either with regard to historical emails/IDs. Which, to be explicit, probably wouldn't happen anytime soon given the effort to impact ratio