perf: Use a global tokio runtime

[andygrove]

Which issue does this PR close?

Closes #1590

Helps with #1523

Rationale for this change

If I configure Spark executor with 8GB + 1GB offHeap and using default settings for tokio threads, q4 hangs in main branch, but completes with these changes.

What changes are included in this PR?

• Allocate one global tokio runtime per process (executor) rather than create one runtime per query
• Use tokio defaults for number of threads
• Allow tokio thread counts to be configured with env vars

Based on an executor configured for 8 concurrent tasks and 1 core per task, the defaults for the overall process now change as follows:

Config	Before	After
worker threads	32	8
max blocking threads	80	512

How are these changes tested?

Manual testing. I do not see any change in overall TPC-H performance, but I can now get q4 to complete with less memory than before.

Note that we cannot close #1523 until apache/datafusion#15323 is resolved.

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 58.55%. Comparing base (f09f8af) to head (099edab).
Report is 127 commits behind head on main.

Additional details and impacted files

@@             Coverage Diff              @@
##               main    #1614      +/-   ##
============================================
+ Coverage     56.12%   58.55%   +2.42%     
- Complexity      976     1063      +87     
============================================
  Files           119      125       +6     
  Lines         11743    12582     +839     
  Branches       2251     2374     +123     
============================================
+ Hits           6591     7367     +776     
- Misses         4012     4020       +8     
- Partials       1140     1195      +55     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[andygrove]

@Kontinuation @wForget could you review?

[Kontinuation]

LGTM. Now we are using environment variables to configure tokio runtime, the proper way of setting up these environment variables on Spark executors will be --conf spark.executorEnv.COMET_BLOCKING_THREADS=N in a non-local setup.

[wForget]

Thank you, LGTM

[andygrove]

@comphead @parthchandra could I get a committer review?

[parthchandra]

lgtm

[andygrove]

Thanks for the reviews @Kontinuation @wForget @parthchandra