Add profiling and throughput reporting to the benchmarks

[alexander-beedie]

While looking for opportunities to improve performance, I extended the benchmarking script with additional profiling and throughput info/options 👍

Three main additions:

• A new kB/s stat/column

  Scaling benchmarks only catch super-linear growth; a uniform slowdown would leave every exponent exactly where it was, bit would show up here.

  $ python benchmarks/bench_parse_throughput.py --vector "UNION ALL"
  UNION ALL chain:
  n input time ratio kB/s status
  100 9885 B 32.2 ms - 307 ok
  200 20.0 kB 62.3 ms 1.94x 321 ok
  400 40.2 kB 136.2 ms 2.19x 295 ok
  800 80.7 kB 317.4 ms 2.33x 254 ok
  scaling exponent 1.17 (1.0 linear, 2.0 quadratic) => linear
  verdict: linear (1 vector, 0 super-linear)
  

• --profile

  Runs each vector once (at its largest size) and print the hottest frames by tottime

  $ python benchmarks/bench_parse_throughput.py --vector "UNION ALL" --profile --profile-top 6
  UNION ALL chain: n=800, 80.7 kB
  ncalls tottime percall cumtime percall filename:lineno(function)
  680760 0.067 0.000 0.067 0.000 {method 'match' of 're.Pattern' objects}
  323171 0.061 0.000 0.125 0.000 sqlparse/utils.py:81(imt)
  47211/11 0.060 0.000 0.160 0.015 sqlparse/engine/grouping.py:478(_group)
  30398 0.056 0.000 0.140 0.000 sqlparse/lexer.py:106(get_tokens)
  60009 0.049 0.000 0.188 0.000 sqlparse/sql.py:235(_token_matching)
  800 0.043 0.000 0.083 0.000 sqlparse/sql.py:310(token_index)
  

• bench_parse_throughput.py

  Covers a selection of large/realistic SQL not found in the other scripts: CTEs, joins, window functions, UNION ALL chains, multi-statement scripts, comment-heavy text.

---

Found several nice speedups using the new flags, for which I'll make separate PRs shortly.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 97.20%. Comparing base (111b35c) to head (1cd995b).
⚠️ Report is 25 commits behind head on master.

Additional details and impacted files

@@ Coverage Diff @@## master #879 +/- ##
==========================================
+ Coverage 97.13% 97.20% +0.07% 
==========================================
Files 31 31 Lines 3663 3793 +130 Branches 328 344 +16 ==========================================
+ Hits 3558 3687 +129 - Misses 63 65 +2 + Partials 42 41 -1 

☔ View full report in Codecov by Harness.
📢 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.