Reduce string allocations in IConsoleOutput implementations

[TheTonttu]

Fixes

• Improves Intermediate heap allocations #2725

Proposed Changes

• Change IConsoleOutput.Write(string) to IConsoleOutput.Write(ReadOnlySpan<char>).
  • More flexible buffer options for caller.
• NetOutput allocation optimizations.
  • Avoid Rune.ToString()
  • Console.Write(object) -> StringBuilder.ToString() vs
    Console.Out.Write(StringBuilder) -> TextWriter.Write(StringBuilder).
• WindowsOutput allocation optimizations.
  • Change WriteConsole Interop to take ReadOnlySpan<char> instead of string.
    • Note that entry point name requires the A or W suffix because LibraryImport does not choose it automatically.
  • Copy to rented array instead of StringBuilder.ToString() during write.
  • Remove unnecessary StringBuilder from SetCursorVisibility.
• Change WindowsConsole WriteConsole Interop to take ReadOnlySpan<char> instead of string.
  • For verifying that this alone does not cause regressions.
• Use EscSeqUtils.CSI_Append(Foreground|Background)ColorRGB in IConsoleOutput implementations.
• Add EscSeqUtils void CSI_WriteCursorPosition(TextWriter, int, int) alternative to string CSI_SetCursorPosition(int, int).
• Change EscSeqUtils.CSI_(Enable|Disable)MouseEvents static properties to readonly fields
  • Consistent with the other similar fields.
  • Cannot accidentally reassign the values on runtime.

Pull Request Checklist

• I've named my PR in the form of "Fixes #issue. Terse description."
• My code follows the style guidelines of Terminal.Gui - if you use Visual Studio, hit CTRL-K-D to automatically reformat your files before committing.
• My code follows the Terminal.Gui library design guidelines
• I ran dotnet test before commit
• I have made corresponding changes to the API documentation (using /// style comments)
• My changes generate no new warnings
• I have checked my code and corrected any poor grammar or misspellings
• I conducted basic QA to assure all features are working

UI Catalog benchmark

No noticeable change to times. Both v2win and v2net finished in ~03:48. I'm guessing the v2 application or drivers have input throttling/pacing which kind of makes the benchmark timings moot for them.

For comparison v1 Windows driver finished in ~00:35.

Microbenchmarks

CSI_SetCursorPosition vs CSI_WriteCursorPosition

BenchmarkDotNet v0.14.0, Windows 10 (10.0.19045.5487/22H2/2022Update)
AMD Ryzen 7 5800X3D, 1 CPU, 16 logical and 8 physical cores
.NET SDK 8.0.406
  [Host]     : .NET 8.0.13 (8.0.1325.6609), X64 RyuJIT AVX2
  DefaultJob : .NET 8.0.13 (8.0.1325.6609), X64 RyuJIT AVX2

writer=Syste(...)riter [35]  

Method	Mean	Error	StdDev	Median	Ratio	RatioSD	Gen0	Allocated	Alloc Ratio
Set	47.04 ns	1.278 ns	3.769 ns	44.46 ns	1.01	0.11	0.0008	40 B	1.00
Write	27.60 ns	0.084 ns	0.065 ns	27.59 ns	0.59	0.04	-	-	0.00

Allocations

v2net

Reduced string allocations nicely.

Before	After

v2win

Not that much improvement. Eliminated SetCursorVisibility char array and StringBuilder allocations.

Before	After

[tznind]

Regarding throttling in v2, I added

    public static ushort MaximumIterationsPerSecond = 25;

In #3949

And set it to ushort max for benchmark.

But it is WIP, benchmark really needs better way of streaming the keys than TimedEvents