Add support for ASGI pathsend extension

[gi0baro]

Summary

As per title, this adds support for ASGI pathsend extension on FileResponse class for servers implementing it.

This is a re-work of #2435 combined with parts of #2616, as discussed in #2613 with @Kludex.

Checklist

• I understand that this PR may be closed in case there was no previous discussion. (This doesn't apply to typos!)
• I've added a test for each change that was introduced, and I tried as much as possible to make a single atomic change.
• I've updated the documentation accordingly.

[gi0baro]

@Kludex I just rebased from current master for your convenience

[gi0baro]

@Kludex what's your current take on this? Should I rebase so it can be merged, or should we just forget about it?

[Kludex]

@Kludex what's your current take on this? Should I rebase so it can be merged, or should we just forget about it?

I want this to get in, I want to get back to this soon. I can rebase myself.

Sorry the delay here.

[Kludex]

This has been going for long - Sorry about it.

The only blocker here is deciding what to do about the changes on the BaseHTTPMiddleware - but I guess that's really the point of this PR. Would it make sense to remove the extension from the scope?

[Kludex]

Is altering the scope["extensions"] in the application side allowed? Would it be bad to remove the http.response.httpsend from the BaseHTTPMiddleware? It's a question. I'm not sure what would be best.

I would like to avoid adding complexity to the BaseHTTPMiddleware.

[gi0baro]

I'm not sure how altering the scope would help in avoiding changes on BaseHTTPMiddleware.

To my perspective, the only way to avoid these lines https://github.com/encode/starlette/pull/2671/files#diff-931c53548e8908544bd0818ee1a8d5f1ada4f07430a439e9322ab369eb40a8acR161-R165 is to re-implement the whole streaming wrapper, which doesn't sound less complex to me tbh..
Maybe we could try to avoid the response re-wrap in _StreamingResponse depending on the context? Probably I don't have the needed to knowledge of Starlette internals to propose such a change though :(

[gi0baro]

@Kludex not sure why the coverage fails after the last rebase..

[Kludex]

@Kludex not sure why the coverage fails after the last rebase..

Because we recently set branch = True on coverage.

[gi0baro]

@Kludex sorry for the long delay.

This is now rebased on latest master, and tests/CI is finally green. Let me know if any additional edits are required for merge.

[Kludex]

Sorry the delay. Gradly you caught me at PyCon IT. 😅