Reduce unsafe usage in TextInfo.cs

[GrabYourPitchforks]

Getting rid of some low-hanging fruit.

The only place this helper is used is from the CultureInfo ctor, via:

runtime/src/libraries/System.Private.CoreLib/src/System/Globalization/CultureData.cs

Lines 2286 to 2289 in c8675d9

	/// <remarks>
	/// This is ONLY used for caching names and shouldn't be used for anything else
	/// </remarks>
	internal static string AnsiToLower(string testString) => TextInfo.ToLowerAsciiInvariant(testString);

Representative call site (see line 1033):

runtime/src/libraries/System.Private.CoreLib/src/System/Globalization/CultureInfo.cs

Lines 1029 to 1033 in c8675d9

	public static CultureInfo GetCultureInfo(string name)
	{
	ArgumentNullException.ThrowIfNull(name);
	name = CultureData.AnsiToLower(name);

This is: (a) not a hot code path; and (b) expected to contain very small strings, typically of the form aa-BB. Seems like it's an ideal candidate for simplifying the code.

[Copilot]

Pull request overview

This PR successfully reduces unsafe code usage in TextInfo.cs by refactoring the ToLowerAsciiInvariant(string) method to use modern, safe C# patterns. The method is used internally for caching culture names (typically short strings like "aa-BB") and is not performance-critical.

Key Changes

• Removed unsafe modifier from the method signature
• Replaced pointer-based operations (fixed, FastAllocateString) with string.Create API
• Simplified the implementation using span-based iteration with foreach (ref char c in span)

[tarekgh]

LGTM!

[MihaZupan]

/ba-g AOT timeouts

[jkotas]

/ba-g AOT timeouts

@MihaZupan NativeAOT failures were not all timeouts:

It is a real failure:

****************************************************
* Size test                                        *
* Size of the executable is   1,555 kB             *
****************************************************
BUG: File size is not in the expected range (1116160 to 1536000 bytes). Did a libraries change regress size of Hello World?
Expected: 100
Actual: 1

I suspect that this regression was introduced by adding a dependency on ValueTuple<T1,T2> that is quite heavy type. I am validating revert in #123692

[jkotas]

MichalStrehovsky/rt-sz#203 confirmed this change introduced 200kB - 300kB regression in hello world NAOT binary size.