fix: Ensure floating-point accuracy in hist

[mcrumiller]

Closes #22234

Two minor adjustments:

1. When user-supplied bins are provided, the slow path is always taken. This guarantees that the user's bins are always respected. The prior implementation attempted to detect if the bins were uniformly distributed using a heuristic, but that heuristic can be foiled, and we were bound to encounter an issue on this sometime in the future.

2. In the fast path (now only when bin_count is supplied), the index calculation might generate results > 1 ULP of the bin edges in some cases, resulting in an invalid index, as see in the linked issue. There, we have 1.0 - 1e-17 == 1.0:

   >>> import polars as pl
   >>> pl.Series([1.0]) - 1e-17 == 1.0
   shape: (1,)
   Series: '' [bool]
   [
           true
   ]

   There is now a check to see whether the value should be moved to an adjacent bucket. This check encompasses a prior check, so performance should not be affected.

[codecov]

Codecov Report

Attention: Patch coverage is 92.30769% with 1 line in your changes missing coverage. Please review.

Project coverage is 80.77%. Comparing base (69ad1ba) to head (e3f710d).
Report is 6 commits behind head on main.

Files with missing lines	Patch %	Lines
crates/polars-ops/src/chunked_array/hist.rs	92.30%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #22245      +/-   ##
==========================================
- Coverage   80.78%   80.77%   -0.01%     
==========================================
  Files        1639     1639              
  Lines      235548   235538      -10     
  Branches     2714     2714              
==========================================
- Hits       190278   190261      -17     
- Misses      44630    44637       +7     
  Partials      640      640              

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.