Correct handling of tagged prereleases

[cameel]

Depends on #16245. Merged.
Depends on #16252. Merged.
Related to #16244.

Currently our scripts do not handle tagged prereleases correctly. They assume that a tag always represents a release, at most filtering out ones not starting with v (but not consistently). This PR adds a distinction between:

• releases: tags consisting of v followed by version number only
• tagged prereleases: tags consisting of v followed by version number and -pre. suffix.
• other prereleases, such as nightlies, CI builds, random builds from develop.

I also took the opportunity to remove some inconsistencies and outdated elements I spotted:

• conf.py in the docs was wrongly assuming that missing prerelease.txt means a release.
• Special-casing of release branch, which we no longer have.
• Unnecessary creation of commit_hash.txt in CI builds.
• Prerelease source tarball always using nightly in its version, regardless of what is in prerelease.txt.

The PR is incomplete, we still need to:

• Update build_win.ps1 to apply the same prerelease logic on Windows.
• Adjust c_release_binaries job to run on tagged prereleases.
• Test that it all works in practice, including:
  • solc-bin properly picking up prerelease binaries as such (by filling out prerelease field in list.json).
  • FORCE_RELEASE being passed all the way down to prerelease_suffix.sh.
  • Scripts which we do not run in CI, e.g. source tarball creation.

[cameel]

Note that there are several bits of prerelease logic that I did not bother to update here:

• docker_deploy_manual.sh and release_ppa.sh both check for a release.
  • They will be removed in Officially discontinue PPA and DockerHub releases #16243 anyway.
• rebuild_current.sh (Scripts used to rebuild old compiler versions to wasm)
• snap/snapcraft.yml (Snap package)
• The CI bits and scripts that use scripts/get_version.sh will get a version without the suffix even on a prerelease tag, but as far as I can tell, that should not cause problems (would be good for someone to double check just in case).

[cameel]

I'm not really sure if there was a reason we were creating commit_hash.txt here, but it seems completely unnecessary now. It appeared a long time ago in #3470, with no explanation, even though the logic for getting it from git already existed in buildinfo.cmake at that time. Then it spread to other places when this logic was copied.

[cameel]

Note that tagged prereleases will skip most of the CI in solc-bin, because compare_bytecode_reports.sh filters out anything that's not a release. I think it's fine, because they are supposed to have passed CI checks in the solidity repo and messing this up is not as big of a deal as with a full release. Still, one thing that would be useful to keep is the sanity check that the filename matches the output of solc --version. This check now sits in validate_reported_version() in bytecode_reports_for_modified_binaries.sh. We should extract it into a separate PR check that runs on all binaries.

[clonker]

Sounds to me like it's almost easier to adapt the regex to allow for prereleases as well:)

[cameel]

Well, that's an option too.

[cameel]

build_win.ps1 has been addressed by converting it to a Bash script (#16252).

c_release_binaries job actually does not need adjustment. Upon a closer look, the regex preserves the prerelease suffix. I tested it for nightlies and I think it will work for -pre releases the same way.

[nikola-matic]

prerelease_file = '../prerelease.txt' would have been nice

[cameel]

I intentionally didn't want to introduce new variables here since I'm not sure if these don't become visible as settings.