Switch NameMangler to UTF-8

[MichalStrehovsky]

Removes some unnecessary UTF-8 -> UTF-16 -> UTF-8 conversions.

Before this change, when compiling hello world: 529000 byte[] allocations, 403000 string allocations.
After this change: 552000 byte[] allocations, 302000 string allocations.

Cc @dotnet/ilc-contrib

[dotnet-policy-service]

Tagging subscribers to this area: @agocke, @MichalStrehovsky, @jkotas
See info in area-owners.md if you want to be subscribed.

[Copilot]

Pull request overview

This PR optimizes the ILCompiler by switching the NameMangler from UTF-16 strings to UTF-8, eliminating unnecessary encoding conversions during compilation. The change reduces string allocations by approximately 101,000 instances when compiling a hello world application (from 403,000 to 302,000), though it increases byte array allocations slightly (from 529,000 to 552,000).

Key Changes:

• Modified NameMangler.GetMangledTypeName() and related methods to return Utf8String instead of string
• Implemented UTF-8 versions of name sanitization and disambiguation logic
• Added UTF-8 string writing overloads in debug info builders and section writers

Reviewed changes

Copilot reviewed 12 out of 12 changed files in this pull request and generated 6 comments.

Show a summary per file

File	Description
src/coreclr/tools/Common/Compiler/NameMangler.cs	Changed abstract method signatures from string to Utf8String for type and name mangling
src/coreclr/tools/Common/Compiler/NativeAotNameMangler.cs	Implemented UTF-8 versions of name mangling, sanitization, and disambiguation logic; added CountDigits helper for efficient numeric formatting
src/coreclr/tools/Common/Internal/Text/Utf8String.cs	Added Concat method to combine multiple UTF-8 strings without conversion
src/coreclr/tools/Common/Internal/Text/Utf8StringBuilder.cs	Added constructor accepting initial capacity parameter
src/coreclr/tools/Common/Compiler/ObjectWriter/SectionWriter.cs	Added WriteUtf8String(Utf8String) overload to write UTF-8 strings directly
src/coreclr/tools/Common/TypeSystem/TypesDebugInfoWriter/TypesDebugInfoWriter.cs	Changed interface methods and type descriptors to use Utf8String
src/coreclr/tools/aot/ILCompiler.Compiler/Compiler/ObjectWriter/Dwarf/DwarfInfoWriter.cs	Added WriteStringReference(Utf8String) overload for UTF-8 string references
src/coreclr/tools/aot/ILCompiler.Compiler/Compiler/ObjectWriter/Dwarf/DwarfBuilder.cs	Changed GetMangledName to return Utf8String
src/coreclr/tools/aot/ILCompiler.Compiler/Compiler/ObjectWriter/CodeView/CodeViewTypesBuilder.cs	Updated user-defined types to use Utf8String; added Write(Utf8String) overload
src/coreclr/tools/aot/ILCompiler.Compiler/Compiler/ObjectWriter/CodeView/CodeViewSymbolsBuilder.cs	Updated WriteUserDefinedTypes signature and added Write(Utf8String) overload
src/coreclr/tools/aot/ILCompiler.Compiler/Compiler/WindowsNodeMangler.cs	Added ToString() call to convert UTF-8 result to string for method table names
src/coreclr/tools/aot/ILCompiler.Compiler/Compiler/UnixNodeMangler.cs	Added ToString() call to convert UTF-8 result to string for method table names

[agocke]

Nice! Do we have any unit tests for the name mangler? If not I could have copilot give it a shot

[MichalStrehovsky]

Nice! Do we have any unit tests for the name mangler? If not I could have copilot give it a shot

I validated this by comparing the output object file (it was byte identical). We don't have unit tests; we can try giving it a shot if copilot comes up with something useful. But name mangling is typically not touched so I can only see the unit test getting in the way if we ever change how we want to mangle things (it is an implementation detail and not a contract). A unit test would definitely make this PR more annoying to do.