build: Remove separate bundle size pipeline

[alexvy86]

Description

Removes the separate pipeline we use in the public project to calculate bundle sizes (and test coverage, which was removed from it by mistake recently), and starts leveraging the main build pipeline of the client release group for that (still in the public project).

This requires that we start running the main build pipeline in the public project for every commit in main, which we don't do today but it's easily configurable (it would replace the corresponding run of the bundle size pipeline; would take longer, but would not be a blocker on the PR gate).

Reviewer Guidance

The review process is outlined on this wiki page.

Note that this does not remove the bundle size pipeline in ADO, just its definition file, so it wouldn't trigger anymore. But as long as we don't delete it in ADO, reverting this PR (and disabling build-client from running for every commit in main in the public project, if we already did that) would get us back to the current state.

[alexvy86]

Bundle analysis and code coverage reports correctly published as artifacts for runs of the Build - client packages: https://dev.azure.com/fluidframework/public/_build/results?buildId=332354&view=artifacts&pathAsName=false&type=publishedArtifacts.

[jatgarg]

@alexvy86 But we haven't merged those PRs yet, right? So the baselines that it would generate won't be correct since we keep updating PRs.

The baseline is not a real baseline in this case.

[alexvy86]

But we haven't merged those PRs yet, right? So the baselines that it would generate won't be correct since we keep updating PRs.

@jatgarg I'm not sure I follow. Which PRs are you referring to? If you mean PRs that are in-flight right now, they do end up in an "intermediate world" but the way I'm thinking about those is that after we merge this, their PR builds will try to find the bundle size / test coverage baselines in runs of build - client packages, won't find one because the build - client packages run for the commit from which those PRs forked off of main wasn't generating baselines yet, and I believe in case of "couldn't find a baseline" we just don't try to calculate deltas nor post a comment to the PR? I'm just not sure if the error would be fatal to the build, but I think it wouldn't.

[microsoft-github-policy-service[bot]]

This PR has been automatically marked as stale because it has had no activity for 60 days. It will be closed if no further activity occurs within 8 days of this comment. Thank you for your contributions to Fluid Framework!