fix: fix performance regression introduced during reader refactor

[westonpace]

In #2913 we added an initialization routine to the scheduler. Unfortunately, this caused a problem. It put schedule_ranges behind an await which means that we wouldn't schedule ranges for all files in a scan immediately. Instead, we wouldn't begin scheduling file X until we had reached nearly the last batch of file X-1. This introduced stuttering in the read path and had a significant effect on performance.

The fix here moves the initialization into the dedicated scheduler thread and keeps the creation of the scheduler a synchronous task which is actually a bit simpler anyways I think.

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 78.26087% with 5 lines in your changes missing coverage. Please review.

Project coverage is 78.02%. Comparing base (d79e870) to head (4be0c48).
Report is 7 commits behind head on main.

Files with missing lines	Patch %	Lines
rust/lance-encoding/src/decoder.rs	78.26%	5 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3170      +/-   ##
==========================================
+ Coverage   77.95%   78.02%   +0.07%     
==========================================
  Files         242      243       +1     
  Lines       82436    82890     +454     
  Branches    82436    82890     +454     
==========================================
+ Hits        64262    64676     +414     
- Misses      14960    15008      +48     
+ Partials     3214     3206       -8     

Flag	Coverage Δ
unittests	78.02% <78.26%> (+0.07%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

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

---

🚨 Try these New Features:

• Flaky Tests Detection - Detect and resolve failed and flaky tests

[westonpace]

Regression for this will be provided by #3165