perf: Reduce decode overhead during pruning keys in the memtable

[evenyag]

I hereby agree to the terms of the GreptimeDB CLA

What's changed and what's your intention?

This PR improves the pruning speed of the new memtable

• Adds decode_value_at() method to decode a value at a specific position in the primary key
  • It will skip values before the position
  • Stores offsets in a buffer so we can reuse them while there are multiple filters to prune
• Adds a PrimaryKeyFilter to filter primary keys via decode_value_at()
• Pushes down filters to Shards and ShardBuilder to reduce DataBatch returned to the partition reader

Checklist

• I have written the necessary rustdoc comments.
• I have added the necessary unit tests and integration tests.
• This PR does not require documentation updates.

Refer to a related PR or issue link (optional)

• Tracking issue for the memtable improvement #2804

[codecov]

Codecov Report

Attention: Patch coverage is 91.91919% with 32 lines in your changes are missing coverage. Please review.

Project coverage is 84.72%. Comparing base (376409b) to head (aea9828).

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3415      +/-   ##
==========================================
- Coverage   85.05%   84.72%   -0.33%     
==========================================
  Files         889      889              
  Lines      146427   146709     +282     
==========================================
- Hits       124547   124303     -244     
- Misses      21880    22406     +526     

[fengjiachun]

LGTM

[v0y4g3r]

LGTM