Reuse existing FormatCallSiteExpression helper for AreEqual/AreNotEqual

[Evangelink]

Follow-up to #8231 addressing #8231 (comment).

The new FormatBinaryCallSiteExpression helper (plus its LineBreakChars field) introduced in #8231 duplicated the existing 2-arg internal FormatCallSiteExpression overload and had subtly different semantics:

• It dropped the entire call-site line when either expression was empty/whitespace, whereas the existing helper only drops it when both are empty and otherwise renders a <placeholder> for the empty side.
• It carried its own LineBreakChars/IndexOfAny newline-detection in parallel with the existing IsMultiline/NormalizeCallSitePlaceholder infrastructure.

This PR:

• Removes FormatBinaryCallSiteExpression and the LineBreakChars field from Assert.cs.
• Routes Assert.AreEqual and Assert.AreNotEqual failure reporting through the existing FormatCallSiteExpression(string, string, string, string, string) overload, matching the call sites already used by Assert.AreSame, Assert.AreNotSame, Assert.Contains, Assert.AreEquivalent, Assert.StartsWith/EndsWith, Assert.MatchesRegex, etc.

Behavioral notes

For the typical case where both [CallerArgumentExpression] values are captured normally, the rendered call site is identical (e.g. Assert.AreEqual(expected, actual)).

The difference only shows up in the rare edge case where one side's caller-argument expression is empty (e.g. used through reflection / hand-built call) - the call site is now still rendered with <expected>/<actual> placeholders, consistent with the other binary assertions, instead of being suppressed.

Validation

• .\build.cmd -c Release: succeeded.
• dotnet test test\UnitTests\TestFramework.UnitTests\TestFramework.UnitTests.csproj -c Release --no-build: only the same 2 pre-existing AreAllDistinct_*WithComparer_HasDuplicate_ShouldFail failures observed on main remain (unrelated to this change).

[Evangelink]

Code Review — PR #8316

✅ 21/21 dimensions clean — no findings.

#	Dimension	Verdict
1	Algorithmic Correctness	✅ LGTM
2	Threading & Concurrency	⏭️ N/A
3	Security & IPC	⏭️ N/A
4	Public API & Binary Compatibility	✅ LGTM
5	Performance & Allocations	✅ LGTM
6	Cross-TFM Compatibility	✅ LGTM
7	Resource & IDisposable Management	⏭️ N/A
8	Defensive Coding at Boundaries	✅ LGTM
9	Localization & Resources	⏭️ N/A
10	Test Isolation	⏭️ N/A
11	Assertion Quality	⏭️ N/A
12	Flakiness Patterns	⏭️ N/A
13	Test Completeness & Coverage	✅ LGTM
14	Data-Driven Test Coverage	⏭️ N/A
15	Code Structure & Simplification	✅ LGTM
16	Naming & Conventions	✅ LGTM
17	Documentation Accuracy	✅ LGTM
18	Analyzer & Code Fix Quality	⏭️ N/A
19	IPC Wire Compatibility	⏭️ N/A
20	Build Infrastructure & Dependencies	⏭️ N/A
21	Scope & PR Discipline	✅ LGTM

Summary

The change is a clean, well-reasoned deduplication:

• Correctness: FormatCallSiteExpression(string,string,string,string,string) correctly handles the new semantics — both-empty → null; one-empty → placeholder; multiline → placeholder. Pre-bracketed literals like "<expected>" pass through NormalizeCallSitePlaceholder unchanged. ✅
• Performance: Marginally better — IsMultiline uses two IndexOf(char) calls (zero-alloc) vs the old IndexOfAny(char[]) which allocated a char[] on pre-.NET 8 targets. ✅
• Cross-TFM: IsMultiline already uses IndexOf(char) explicitly (not string.Contains(char)) for netstandard2.0/net462 compatibility. ✅
• Test coverage: AssertTests.FormatCallSiteExpressionTests.cs covers the behavior-changing edge case (one-empty-one-non-empty → placeholder, not null). ✅
• API compat: Both removed members were private static — no public surface affected. ✅

Generated by Expert Code Review (on open) for issue #8316 · ● 10.9M