Not compatible with `--noexperimental_python_import_all_repositories`

The `--noexperimental_python_import_all_repositories` setting is useful in working around https://github.com/bazelbuild/bazel/issues/14640 with the PYTHONPATH becoming too large if you have python targets with large numbers of dependencies.

But `rules_apple` does not seem to be compatible with this. Specifically, when using `apple_core_ml_library` with `--noexperimental_python_import_all_repositories`

```
ERROR:*****/BUILD:346:22: MlmodelGenerate ****/****.m failed: (Exit 1): xctoolrunner failed: error executing MlmodelGenerate command (from target //****:****.CoreML) bazel-out/darwin_arm64-opt-exec-ST-13d3ddad9198/bin/external/build_bazel_rules_apple/tools/xctoolrunner/xctoolrunner coremlc generate ... (remaining 2 arguments skipped)

Use --sandbox_debug to see verbose messages from the sandbox and retain the sandbox build root for debugging

Traceback (most recent call last):

File "****/../build_bazel_rules_apple/tools/xctoolrunner/xctoolrunner.py", line 43, in

from tools.wrapper_common import execute

ModuleNotFoundError: No module named 'tools'
```

Start with tools/xctoolrunner/xctoolrunner.py and the apple_core_ml_library implementation, then reproduce the failure using --noexperimental_python_import_all_repositories. Trace why the xctoolrunner invocation cannot import tools; done means the Core ML target builds successfully with that setting enabled.

Written by the indexing model from the issue text.