Work around some issues where recording streams leaking context when used with generators

[jleibs]

What

• Resolves: RecordingStream context managers are broken when used with generators  #6238

In thread_local_stream, don't yield while holding the stream context open. Re-create context before continuing the generator.

Also introduce a new recording_stream_generator_ctx for more advanced usage. This is mainly an escape hatch.

The problem from the example can now be handled using:

import rerun as rr

@rr.recording_stream_generator_ctx
def my_gen_func(stream, name):
    with stream:
        for i in range(10):
            print(f"{name} {i}")
            rr.log("stream", rr.TextLog(f"{name} {i}"))
            yield

rr.init("rerun_example_leak_context")

stream1 = rr.new_recording("rerun_example_stream1")
stream1.save("stream1.rrd")
stream2 = rr.new_recording("rerun_example_stream2")
stream2.save("stream2.rrd")

gen1 = my_gen_func(stream1, "stream1")
gen2 = my_gen_func(stream2, "stream2")

next(gen1)
next(gen2)
rr.log("stream", rr.TextLog("This should go to the global stream"))
next(gen1)
next(gen1)
next(gen1)

Checklist

• I have read and agree to Contributor Guide and the Code of Conduct
• I've included a screenshot or gif (if applicable)
• I have tested the web demo (if applicable):
  • Using examples from latest main build: rerun.io/viewer
  • Using full set of examples from nightly build: rerun.io/viewer
• The PR title and labels are set such as to maximize their usefulness for the next release's CHANGELOG
• If applicable, add a new check to the release checklist!

• PR Build Summary
• Recent benchmark results
• Wasm size tracking

To run all checks from main, comment on the PR with @rerun-bot full-check.

[abey79]

Great stuff! Extensively discussed in a huddle. TLDR: some more comments welcome and the __enter__ call in the finally clause of recording_stream_generator_ctx needs revisiting.