Refresh benchmarks against current main

[pscheid92]

Regenerates docs/BENCHMARKS.md on the current tree (Apple M2, macOS 26, .NET 10, 2026-07-23). The committed numbers predated the IAeadCipher backend seam and the pre-auth parser bounds.

What changed

• CLI comparison: re-run against a freshly-built current AOT binary. Essentially unchanged — within wall-clock noise, and every relative ranking holds (Go leads binary 100 MB; age-sharp beats rage there; age-sharp beats both on armored 100 MB encrypt).
• Microbenchmarks: times moved a few percent. The one substantive change is per-operation allocation on the ChaCha path (Encrypt 1 KB ~12→20 KB, X25519 Wrap ~3.6→6.6 KB), attributable to the cipher seam allocating a backend per operation. Streaming/per-chunk behavior is unchanged (Random Access allocations flat), so the O(1)-memory guarantee for large files still holds.
• The armored-decrypt microbenchmark — the only hot path the new parser bounds touch — stayed within noise, confirming the bound's cost is negligible.

Docs-only; no code change.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 91.32%. Comparing base (88ba9b8) to head (6c4a1f8).
⚠️ Report is 2 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #21      +/-   ##
==========================================
+ Coverage   91.28%   91.32%   +0.04%     
==========================================
  Files          42       42              
  Lines        2443     2443              
  Branches      322      322              
==========================================
+ Hits         2230     2231       +1     
+ Misses        154      153       -1     
  Partials       59       59              

☔ 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.