v2.2: Optimize AppendVec / BufferedReader buffer sizes and alloc strategy (backport of #5457)

[mergify]

Problem

BufferedReader instances allocate significantly more memory than is optimal. This creates significant dropping pressure and slows down file reading from disk.

Summary of Changes

This adjusts BufferedReader instances to use 4KiB stack buffers for reads that don't require account data. The instance that does require account data starts with a 1MiB heap buffer and will resize once to worst case (~10MiB) account data size if an account is encountered that doesn't fit in the 1MiB buffer. This is optimized for the typical case where most accounts are token accounts of roughly ~300b.

BufferedReader had to be made generic to support both stack and heap allocated buffers, which adds a small amount of complexity, but gives us the significant advantage of not having to heap allocate every instance.

---

This is an automatic backport of pull request #5457 done by [Mergify](https://mergify.com).

[alessandrod]

why backport: file-io is active in 2.2, we're making it faster

[t-nelson]

@mergify rebase

[mergify]

rebase

✅ Branch has been successfully rebased

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 83.3%. Comparing base (fcba40a) to head (aec8616).

Additional details and impacted files

@@           Coverage Diff           @@
##             v2.2    #5487   +/-   ##
=======================================
  Coverage    83.3%    83.3%           
=======================================
  Files         800      800           
  Lines      370256   370299   +43     
=======================================
+ Hits       308537   308624   +87     
+ Misses      61719    61675   -44     

🚀 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.