Optimize conversion of ModifierKeys enum from/to string, reduce allocations

[h3xds1nz]

Fixes #7502

Description

Parsing optimization for ModifierKeys and its conversion from/to string. Conversion from is now alloc-free, conversion to is either free or reduced by a big margin (see benchmarks).

• ConvertFrom is now fully flying on ReadOnlySpan<char> instead of multiple substrings and trims on string instances.
• ConvertTo is now within 2 procs, fast-path and slow-path. The only allocated string instance is with multiple modifiers.
• Strings in ConvertTo will now be interned with the rest since we use StringComparison.OrdinalIgnoreCase.
• ModifierKeysFlag static field is now a compile-time constant, saving memory and providing speed-up.
• The JITTed codegen takes up less memory as we removed few things.
• Adds/completes code documentation for the whole class.

ConvertFrom Benchmark for all possible modifiers

Method	source	Mean [ns]	Error [ns]	StdDev [ns]	Gen0	Code Size	Allocated
Original	CTRL (...)+ ALT [24]	101.55 ns	1.790 ns	1.587 ns	0.0224	2,662 B	376 B
PR__EDIT	CTRL (...)+ ALT [24]	34.68 ns	0.218 ns	0.182 ns	-	1,891 B	-
Original	Ctrl (...)+ Alt [28]	147.59 ns	1.616 ns	1.512 ns	0.0319	2,640 B	536 B
PR__EDIT	Ctrl (...)+ Alt [28]	35.19 ns	0.483 ns	0.451 ns	-	1,891 B	-

ConvertTo benchmark for different scenarios

Method	value	Mean [ns]	Error [ns]	StdDev [ns]	Gen0	Code Size [B]	Allocated [B]
Original	None	1.0984 ns	0.0108 ns	0.0090 ns	-	685 B	-
PR__EDIT	None	0.8665 ns	0.0078 ns	0.0066 ns	-	238 B	-
Original	Shift	1.1785 ns	0.0149 ns	0.0132 ns	-	780 B	-
PR__EDIT	Shift	0.8840 ns	0.0209 ns	0.0196 ns	-	255 B	-
Original	3	10.517 ns	0.1322 ns	0.1104 ns	0.0043	1,134 B	72 B
PR__EDIT	3	9.546 ns	0.1713 ns	0.1518 ns	0.0024	756 B	40 B
Original	15	39.23 ns	0.828 ns	0.886 ns	0.0176	1,137 B	296 B
PR__EDIT	15	11.88 ns	0.194 ns	0.181 ns	0.0043	753 B	72 B

Customer Impact

Increased performance, zero/reduced allocations, less static memory.

Regression

No.

Testing

Local build, CI, assert tests:

//ConvertTo
string x1 = Original(ModifierKeys.None);
string x2 = PR__EDIT(ModifierKeys.None);
Assert.AreEqual(x1, x2);
x1 = Original(ModifierKeys.Shift);
x2 = PR__EDIT(ModifierKeys.Shift);
Assert.AreEqual(x1, x2);
x1 = Original(ModifierKeys.Shift | ModifierKeys.Alt);
x2 = PR__EDIT(ModifierKeys.Shift | ModifierKeys.Alt);
Assert.AreEqual(x1, x2);
x1 = Original(ModifierKeys.Control | ModifierKeys.Alt);
x2 = PR__EDIT(ModifierKeys.Control | ModifierKeys.Alt);
Assert.AreEqual(x1, x2);
x1 = Original(ModifierKeys.Windows | ModifierKeys.Alt);
x2 = PR__EDIT(ModifierKeys.Windows | ModifierKeys.Alt);
Assert.AreEqual(x1, x2);
x1 = Original(ModifierKeys.Windows | ModifierKeys.Alt | ModifierKeys.Control | ModifierKeys.Shift);
x2 = Original(ModifierKeys.Windows | ModifierKeys.Alt | ModifierKeys.Control | ModifierKeys.Shift);
Assert.AreEqual(x1, x2);

//ConvertFrom
ModifierKeys x1 = Original("CTRL + ");
ModifierKeys x2 = PR__EDIT("CTRL + ");
Assert.AreEqual(x1, x2);
x1 = Original("CTRL");
x2 = PR__EDIT("CTRL");
Assert.AreEqual(x1, x2);
x1 = Original(" ALT + SHIFT ");
x2 = PR__EDIT(" ALT + SHIFT ");
Assert.AreEqual(x1, x2);
x1 = Original("WIN + WINDOWS");
x2 = PR__EDIT("WIN + WINDOWS");
Assert.AreEqual(x1, x2);
x1 = Original("   ");
x2 = PR__EDIT("   ");
Assert.AreEqual(x1, x2);
x1 = Original("CTRL + WIN + SHIFT + ALT");
x2 = PR__EDIT("CTRL + WIN + SHIFT + ALT");
Assert.AreEqual(x1, x2);
x1 = Original("CTRL ++ WIN");
x2 = PR__EDIT("CTRL ++ WIN");
Assert.AreEqual(x1, x2);
x1 = Original("ALT+WIN");
x2 = PR__EDIT("ALT+WIN");
Assert.AreEqual(x1, x2);

Risk

Low.

Microsoft Reviewers: Open in CodeFlow

[harshit7962]

@h3xds1nz, the PR can be taken in after resolution of the conflict.

[harshit7962]

Also, as we already have WindowsBase project already set-up for unit tests, can we add some tests for this change?

[h3xds1nz]

@harshit7962 I didn't PR any as there are tests in #8215, same for KeyConverter. I've unfortunately found out after I wrote them myself, haha.

On my way with the conflicts.

[harshit7962]

Reapproving post resolution of conflict.

[harshit7962]

Thanks for the contribution @h3xds1nz.

[h3xds1nz]

@harshit7962 Thanks for taking it :)