perf: Eliminate LINQ allocations and improve algorithmic complexity in Chart segments

[PaulAndersonS]

Root Cause of the Issue

Multiple Chart segment classes use LINQ chains (.Where(), .DefaultIfEmpty(), .Min(), .Max(), .ToList(), .Sum()) in data processing and layout methods. These allocate enumerators, delegate objects, and intermediate collections on every call — causing unnecessary GC pressure during data binding and rendering.

Additionally, CircularSeries.LeftPoints.Contains() performs O(n) linear searches during label arrangement, leading to O(n²) complexity with many data points.

Description of Change

Five targeted performance optimizations:

1. ErrorBarSegment.SetData — Replaced 4 chained LINQ Where/Min/Max calls with a single-pass for loop that computes min/max in one traversal. Eliminates 4 enumerator allocations per data bind.

2. ErrorBarSegment.GetSdErrorValue — Replaced Where().ToList(), two List<double> allocations (dev, sQDev), and Sum(x => x) with two simple loops computing mean and sum-of-squared-deviations in-place. Eliminates 3 List allocations and 1 delegate allocation per call.

3. CircularSeries label arrangement — Added HashSet<PieSegment> LeftPointsLookup for O(1) Contains() checks, replacing O(n) List.Contains() that caused O(n²) behavior during label overlap resolution.

4. AreaSegment.SetData — Replaced LINQ Where/DefaultIfEmpty/Min chain with a simple for loop to find the minimum non-NaN value.

5. SplineAreaSegment.UpdateRange & SplineRangeAreaSegment.UpdateRange — Same LINQ-to-loop optimization for NaN-filtered minimum calculation.

Issues Fixed

N/A — Proactive performance improvement

Test Coverage

• Added 5 new unit tests validating ErrorBarSegment and AreaSegment SetData behavior with NaN and mixed values
• All 2353 existing Chart unit tests pass without regression

Screenshots

N/A — No visual changes