Mpi4jax

What's Changed
* Do not use thread-local state (fixes 139) by dionhaefner in https://github.com/mpi4jax/mpi4jax/pull/140

**Full Changelog**: https://github.com/mpi4jax/mpi4jax/compare/v0.3.7...v0.3.8

What's Changed
* Add missing barrier primitive registration on GPU by dionhaefner in https://github.com/mpi4jax/mpi4jax/pull/118
* Created auto_tokenize by Thenerdstation in https://github.com/mpi4jax/mpi4jax/pull/111

New Contributors
* Thenerdstation made their first contribution in https://github.com/mpi4jax/mpi4jax/pull/111

**Full Changelog**: https://github.com/mpi4jax/mpi4jax/compare/v0.3.6...v0.3.7

What's Changed
* Restore support for Mac ARM
* Only retrieve dtype object when needed (fixes 112) by dionhaefner in https://github.com/mpi4jax/mpi4jax/pull/113

**Full Changelog**: https://github.com/mpi4jax/mpi4jax/compare/v0.3.5...v0.3.6

What's Changed
* Fix inconsistent output shapes by dionhaefner in https://github.com/mpi4jax/mpi4jax/pull/110
* Now supports `jax >= 0.2.27`.


**Full Changelog**: https://github.com/mpi4jax/mpi4jax/compare/v0.3.4...v0.3.5

What's Changed
* Adds `mpi4jax.has_cuda_support` to probe if mpi4jax was built with CUDA support


**Full Changelog**: https://github.com/mpi4jax/mpi4jax/compare/v0.3.3...v0.3.4

What's Changed
* Compatibility with jax>=0.2.26 by dionhaefner in https://github.com/mpi4jax/mpi4jax/pull/108


**Full Changelog**: https://github.com/mpi4jax/mpi4jax/compare/v0.3.2...v0.3.3