Please use multithreaded xz

My team are constantly complaining that llvm takes over a minute to extract, even if the download is cached. This is due to bazel using single-threaded java to extract .xz archives. It's possible to use an external xz toolchain which uses multithreading, see https://github.com/google-ml-infra/rules_ml_toolchain/commit/50a9847d5fdbca9d8c0351f5a879bfabd6598ad7 for examples

Start by reviewing the referenced rules_ml_toolchain commit, then locate how this repository configures Bazel extraction of .xz archives. The work is complete when .xz extraction uses an external multithreaded xz toolchain instead of single-threaded Java and the relevant behavior is verified.

Written by the indexing model from the issue text.