PackReader.read(): return a memoryview of the in-memory pack instead of copying

[mr-raj12]

Description

PackReader.read() read chunks from the in-memory pack by slicing pack_contents, a bytes object, so every read allocated and copied. It now returns a memoryview slice, which is a view into the pack with no copy. On a profiled 6 GiB extract that copy cost about 1.3% of extract CPU.

Supporting changes:

• parse() and parse_meta() take bytes or memoryview for cdata (both wrap it in memoryview() anyway).
• get(read_data=False) returns header plus meta as bytes.

The store range-read path is unchanged and still returns bytes. Re-put and key detection already accept the buffer protocol.

fixes #10030

Checklist

• PR is against master
• New code has tests and docs where appropriate
• Tests pass (repository, repoobj, and extract/check/debug consumer subsets)
• Commit messages are clean and reference related issues

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 86.15%. Comparing base (2fe079e) to head (7f0f3cb).
⚠️ Report is 1 commits behind head on master.
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@           Coverage Diff           @@
##           master   #10037   +/-   ##
=======================================
  Coverage   86.15%   86.15%           
=======================================
  Files          96       96           
  Lines       17360    17360           
  Branches     2659     2659           
=======================================
  Hits        14956    14956           
  Misses       1664     1664           
  Partials      740      740           

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