Write extent tombstones only for extents that exist

[GlazingAlgorithm]

Hi Pierre, great project!

I recently encountered an issue while trimming a large block device on NBD. Currently, a trim writes one tombstone per 32k extent in the requested range, regardless of whether the extent has a key. Because Bloom filters count every row, SSTs become bloated with filter blocks of hundreds of MBs when trimming large sparse devices.

File-range deletes are affected by the same issue.

In this fix we reuse the existing range scan (used for debiting) to determine which extents actually need to be deleted.

Thank you for your work!

[cla-assistant]

All committers have signed the CLA.

[cla-assistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 83.24%. Comparing base (5377066) to head (d09b5ed).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #625      +/-   ##
==========================================
+ Coverage   83.23%   83.24%   +0.01%     
==========================================
  Files         104      104              
  Lines       47003    47048      +45     
==========================================
+ Hits        39122    39167      +45     
  Misses       7881     7881              

Files with missing lines	Coverage Δ
zerofs/src/db.rs	86.11% <100.00%> (+0.05%)	⬆️
zerofs/src/fs/store/extent/write.rs	94.48% <100.00%> (+0.15%)	⬆️

... and 3 files with indirect coverage changes

🚀 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.

[Barre]

Looks great, thanks.