Hard-wire --experimental_sibling_repository_layout to false.

[lberki]

trafficstars

According to the discussion https://github.com/bazelbuild/bazel/discussions/20500, no one is too attached to it and the project, sadly, failed: given the current climate, it doesn't look like we'll ever have enough political will to go through with the necessary migration to flip the default value of the flag, as good an idea as it is on its own merits.

This change is designed to minimize the number of lines changed so that it's easy to roll back if need be.

[lberki]

In a last, desperate attempt, I ran the HEAD + downstream projects pipeline and it's surprisingly green: https://buildkite.com/bazel/bazel-at-head-plus-downstream/builds/3753#018e7f5a-9256-492e-8dcf-ccbe3e587943

What fails is the following:

• Genuine domain-specific issues:
  • Bazel (in gathering BUILD.bazel file for @bazel_tools, I believe, seems fixable)
  • IntelliJ plugin (seems to assume the external/ path segment)
  • rules_dotnet (hard-wires assumptions about paths that wouldn't hold anymore)
  • rules_haskell (HaskellBuildLibrary is broken)
  • rules_jvm_external (due to the change to the layout of the runfiles tree)
  • rules_perl: actually crashes Bazel
  • rules_python: genuine breakage, but seems to be fixable
  • rules_foreign_cc (a test that checks an exact path that changes + "file not found issues")
  • rules_testing: an issue in Skydoc
  • rules_webtesting: an issue in Skydoc
• C++ issues:
  • Bazel examples
  • Cloud Robotics (it says "missing input file @llvm_toolchain//:bin/llvm-dwp, so it looks fixable in O(1) time)
  • rules_go (zlib didn't compile -- this is alarming, if zlib didn't compile what else is broken although it might just be .d file parsing issue that could be fixed in Bazel)
  • Protobuf ("this rule is missing dependency declarations", so may be an issue with .d file parsing)
• Probably false positives:
  • rules_kotlin (I fail to see how "No such package @rules_kotlin//kotlin might be caused by this, but maybe my imagination is not rich enough)
  • Skylib (looked like an artifact of how I ran the experiment)
  • Kythe (an unrelated Java failure, may be red at HEAD, too)
  • rules_graalvm (seems very unrelated, I didn't check if it's green at HEAD but I'd no be surprised if it was red)

[comius]

@lberki Is this ready for a review? Does it need 3 reviewers? Can you import it? I find internal reviews more pleasant than GitHub.

[lberki]

If all we do is internal reviews, we are not really an open source project, are we?

My understanding is that we decided to flip the flag instead and start a Gardening project to fix the downstream projects so I'll close this pull request without merging it.