Add byte buffer to improve gremlin-python performance

[GumpacG]

Description

The streaming-support change in #3396 regressed the python CI job from ~5-6 min to 10-19 min and started hitting the 20-min timeout. The root cause is that AiohttpSyncStream.read(n) was running a full asyncio event-loop turn (loop.run_until_complete(...)) for every byte-level read the GraphBinary deserializer performs. A typical response of N results triggers at minimum 3N + 8 reads - for a 5,000-result response that's ~500 ms of pure event-loop overhead vs ~0.5 ms before streaming.

This PR keeps the streaming design intent but pays the asyncio cost at chunk granularity (~64 KB) instead of per byte.

Assisted-by: Devin: Claude Opus 4.7

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 76.37%. Comparing base (cfd6889) to head (6345172).
⚠️ Report is 1050 commits behind head on master.

Additional details and impacted files

@@             Coverage Diff              @@
##             master    #3422      +/-   ##
============================================
- Coverage     77.87%   76.37%   -1.50%     
+ Complexity    13578    13446     -132     
============================================
  Files          1015     1013       -2     
  Lines         59308    60417    +1109     
  Branches       6835     7091     +256     
============================================
- Hits          46184    46142      -42     
- Misses        10817    11557     +740     
- Partials       2307     2718     +411     

☔ 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.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[Cole-Greer]

VOTE +1

[kenhuuu]

VOTE +1

[xiazcy]

VOTE +1