Use char overloads for single-char string operations

[ivandrofly]

Summary

• Replace StartsWith/EndsWith/IndexOf calls taking a single-character string plus StringComparison.Ordinal with the equivalent char overloads (e.g. s.EndsWith("\"", StringComparison.Ordinal) → s.EndsWith('"'))
• Pure refactor across 31 files in libse, ui, and tests — no behavior change

Benchmarks

BenchmarkDotNet 0.14.0, .NET 10, Release, [MemoryDiagnoser], 15 iterations per build — this branch vs parent commit 2131a6b, exercising the changed methods and an aggregate round-trip over the 20 changed subtitle formats (100-paragraph subtitle; 19 exercisable via ToText, 18 via LoadSubtitle):

Benchmark	Time before	Time after	Change
Utilities.ReverseStartAndEndingForRightToLeft	623.7 ns	525.8 ns	−16%
MoveWordUpDown.MoveWordUp	265.8 ns	235.8 ns	−11%
MoveWordUpDown.MoveWordDown	438.4 ns	395.1 ns	−10%
ContinuationUtilities.IsFullLineTag	3.39 µs	3.11 µs	−8%
All 18 formats LoadSubtitle (aggregate)	19.05 ms	17.46 ms	−8%
RemoveTextForHI.RemoveTextFromHearImpaired	6.31 µs	5.86 µs	−7%
All 19 formats ToText (aggregate)	2.32 ms	2.19 ms	−6%
Formatting.SetTagsAndReturnTrimmed + ReAddFormatting	772.9 ns	739.1 ns	−4%

Allocations are byte-identical on every benchmark, as expected — the char overloads change comparison cost, not allocation. Every benchmark moved in the same direction (4–16% faster); the small per-call savings compound because these calls sit inside per-line/per-character parsing loops.

Test plan

• dotnet test tests/libse/LibSETests.csproj passes (745/745 locally)
• dotnet test tests/UI/UITests.csproj passes (913 passed, 1 skipped locally)
• dotnet test tests/libuilogic/LibUiLogicTests.csproj passes (18/18 locally)
• Load a SAMI/JSON/CSV subtitle file and confirm parsing is unchanged

🤖 Generated with Claude Code

[ivandrofly]

Updated @niksedk