remaining_bytes aligns len since all writes will align first

[jeffwashington]

Problem

ancient append vec appending relies on a reliable count of remaining bytes. It turns out that the append vec's len() is up to the last written byte. This can be non-u64 aligned. The result is if we are to append more to the file, the first operation will be a u64_align!. This means that remaining_bytes was reporting too many bytes. Not all those bytes could be written to, thanks to the bytes wasted for alignment. The result was that writes could fail, causing an attempt at creating a second append vec for the slot. This is no longer allowed and will fail with an assert.

Summary of Changes

Align len() so that remaining bytes returns an accurate number.

Fixes #

[codecov]

Codecov Report

Merging #34171 (2f84b89) into master (3368579) will decrease coverage by 0.1%.
Report is 2 commits behind head on master.
The diff coverage is 95.6%.

Additional details and impacted files

@@            Coverage Diff            @@
##           master   #34171     +/-   ##
=========================================
- Coverage    81.9%    81.9%   -0.1%     
=========================================
  Files         819      819             
  Lines      220168   220189     +21     
=========================================
- Hits       180441   180428     -13     
- Misses      39727    39761     +34     

[HaoranYi]

Since we are fixing remaining bytes, has_space_available should be fixed too, stay consistent. right?

[brooksprumo]

lgtm