perf: Optimize xrefmap.json file deserialization performance

[filzrev]

What's changed in this PR

When XrefMap is loaded from local file or network.
The current implementation using YamlDotNet's deserialize method.

For JSON file deserialization.
It seems to be faster to use Newtonsoft.json for JSON deserialization.

Benchmark

YamlDotnet	8781[ms]
Newtonsoft.Json	2304[ms]
System.Text.Json	1687[ms]	Note: It's not included in this PR. I'll create separated PR later.

[codecov]

Codecov Report

Attention: Patch coverage is 54.54545% with 5 lines in your changes are missing coverage. Please review.

Project coverage is 79.02%. Comparing base (fe673ec) to head (062ab3c).
Report is 97 commits behind head on main.

Files	Patch %	Lines
src/Docfx.Build/XRefMaps/XRefMapDownloader.cs	54.54%	3 Missing and 2 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #9824      +/-   ##
==========================================
+ Coverage   74.31%   79.02%   +4.71%     
==========================================
  Files         536      536              
  Lines       23189    23195       +6     
  Branches     4056     4029      -27     
==========================================
+ Hits        17234    18331    +1097     
+ Misses       4853     3740    -1113     
- Partials     1102     1124      +22     

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

[paulushub]

Is file-based database (like sqlite) possible as an option? query might be faster.

[filzrev]

Is file-based database (like sqlite) possible as an option? query might be faster.

I though faster xrefmap loading is important.
Because loading/deserializing xrefmap over 100MB size JSON file take some times and take memory footprint.
And this operation is executed on every docfx build.

So it might be better to use file-based database or some key-value storage as suggested.

Some of the required characteristics are.

1. File-based database/storage.
2. It can use Compressed Database file and that size should be lower than 100MB for .NET xrefmap.
   (100MB is GitHub file size limit without using Git LFS)
   (When compressing .NET xrefmap.json with gzip. It can reduce file size from 131MB to 8.6MB)

[yufeih]

Nice, looking forward to the switch to S.T.J.