feat: batch check updates

[booniepepper]

Description

• feat(client): run sync batch_check with ThreadPoolExecutor for parallelism
• feat(client): use asyncio semaphore for streaming batch check

References

Generated by openfga/sdk-generator#228

Review Checklist

• I have clicked on "allow edits by maintainers".
• I have added documentation for new/changed functionality in this PR or in a PR to openfga.dev [Provide a link to any relevant PRs in the references section above]
• The correct base branch is being used, if not main
• I have added tests to validate that the change in functionality is working as expected

[booniepepper]

TODO: Do some performance testing on this vs the async client

[booniepepper]

https://github.com/booniepepper/openfga-python-batch-check-bench

"core" branch has tests against api.us1.fga.dev and the "local" branch expects fga server running on local host (i.e. from docker)

Here are some excerpts of test output using https://github.com/sharkdp/hyperfine:

Benchmark 1: async
  Time (mean ± σ):     505.1 ms ± 170.1 ms    [User: 359.1 ms, System: 99.6 ms]
  Range (min … max):   404.0 ms … 1967.3 ms    100 runs

Benchmark 2: async+semaphore
  Time (mean ± σ):     443.2 ms ±  19.0 ms    [User: 329.3 ms, System: 88.6 ms]
  Range (min … max):   403.7 ms … 493.1 ms    100 runs

Benchmark 3: sync+threadpool
  Time (mean ± σ):     431.1 ms ±  15.6 ms    [User: 319.9 ms, System: 87.8 ms]
  Range (min … max):   408.7 ms … 495.2 ms    100 runs

So with my changes here I see about a 20% performance gain on OpenFgaClient.batch_check(...) performance

Edit: The benchmark was flawed at this time. It was using CheckRequest instead of ClientCheckRequest which failed to actually call the underlying check() functions

[booniepepper]

For posterity benchmarks lead to these results:

Benchmark: async (current)
  Time (mean ± σ):      1.096 s ±  0.051 s    [User: 0.752 s, System: 0.171 s]
  Range (min … max):    1.029 s …  1.482 s    1000 runs

Benchmark: async+semaphore+gather
  Time (mean ± σ):      1.022 s ±  0.044 s    [User: 0.743 s, System: 0.170 s]
  Range (min … max):    0.957 s …  1.277 s    1000 runs

Benchmark: sync+threadpool
  Time (mean ± σ):     962.3 ms ±  38.3 ms    [User: 785.2 ms, System: 208.9 ms]
  Range (min … max):   896.5 ms … 1087.3 ms    100 runs

Where "current" was from the commit prior to commits in this PR.