Optimize CSS parser performance

[jafin]

Prerequisites

Please make sure you can check the following two boxes:

• I have read the CONTRIBUTING document
• My code follows the code style of this project

Contribution Type

What types of changes does your code introduce? Put an x in all the boxes that apply:

• Bug fix (non-breaking change which fixes an issue, please reference the issue id)
• New feature (non-breaking change which adds functionality, make sure to open an associated issue first)
• Breaking change (fix or feature that would cause existing functionality to change)
• My change requires a change to the documentation
• I have updated the documentation accordingly
• I have added tests to cover my changes
• All new and existing tests passed

Description

• Refactor ContentFrom() to scan chars directly instead of re-invoking the full tokenizer, eliminating double-tokenization
• Cache single-char token strings via static lookup table.
• Add ToLowerFast() to skip allocation when strings are already lowercase, use StringBuilder for custom property name concatenation
• Replace @-rule if-else chain with dictionary dispatch, add dictionary index to CssStyleDeclaration for O(1) property lookups
• Avoid double-array allocation in PeriodicValueConverter when all 4 values are present

CSS File	Before	After	Speedup	Mem Before	Mem After	Mem Saved
cdnjs.cloudflare	2.21 ms	1.95 ms	1.13x	1.99 MB	1.81 MB	9%
csszengarden	2.35 ms	1.82 ms	1.29x	1.84 MB	1.63 MB	11%
florian-rappl	4.73 ms	4.19 ms	1.13x	3.47 MB	3.34 MB	4%
maxcdn.bootstrapcdn	8.66 ms	7.69 ms	1.13x	6.69 MB	5.72 MB	15%
s.yimg	2.02 ms	1.91 ms	1.06x	2.83 MB	2.50 MB	12%
static.licdn	2.28 ms	2.09 ms	1.09x	2.27 MB	2.03 MB	11%
style.aliunicorn	1.59 ms	1.54 ms	1.03x	2.17 MB	2.04 MB	6%
z-ecx.images-amazon	8.46 ms	8.06 ms	1.05x	6.85 MB	6.05 MB	12%

AngleSharp vs ExCss comparison (after optimizations)

CSS File	AngleSharp	ExCss	Ratio	Memory Ratio
cdnjs.cloudflare	1.95 ms	2.24 ms	0.87x	2.00x less
csszengarden	1.82 ms	1.43 ms	1.28x	1.35x less
florian-rappl	4.19 ms	11.57 ms	0.36x	4.26x less
maxcdn.bootstrapcdn	7.69 ms	9.56 ms	0.80x	2.03x less
s.yimg	1.91 ms	11.25 ms	0.17x	5.63x less
static.licdn	2.09 ms	6.10 ms	0.34x	3.89x less
style.aliunicorn	1.54 ms	10.76 ms	0.14x	6.83x less
z-ecx.images-amazon	8.06 ms	10.45 ms	0.77x	2.09x less

Summary:
AngleSharp is now faster on 7 of 8 benchmark files and allocates 2–7× less memory than ExCss across all files.

[jogibear9988]

ToLowerFast maybe still could be optimized:

https://dotnetfiddle.net/ToLmEm

[jafin]

@jogibear9988 I'll take a look and see what the benchmark says.
I'll report back with the results of the ToLower changes you've posted would be. If it's a win I'll amend my PR.

[jafin]

Scenario	Method	Run 1	Run 2	Allocated
Already lowercase	Original	7.66 ns	7.62 ns	0 B
	Span	4.29 ns	4.26 ns	0 B
	SIMD	8.01 ns	8.19 ns	0 B
Mixed case	Original	20.18 ns	20.00 ns	56 B
	Span	19.07 ns	19.49 ns	56 B
	SIMD	20.12 ns	20.22 ns	56 B
All uppercase	Original	14.16 ns	14.37 ns	56 B
	Span	15.05 ns	15.00 ns	56 B
	SIMD	15.25 ns	15.21 ns	56 B

Both runs confirm the same pattern:

• Span is ~44% faster for the case that matters most since CSS identifiers are mostly lowercase
• Mixed/uppercase cases: negligible difference (dominated by ToLowerInvariant allocation)
• SIMD adds overhead for these short strings.
• Zero regressions in the full test suite

CSS File Benchmark Comparison Existing PR vs adjusting to Span

CSS File	Before (Original)	After (Span)	Delta
cdnjs.cloudflare	42ms	43ms	+1ms
csszengarden	10ms	11ms	+1ms
florian-rappl	22ms	21ms	-1ms
maxcdn.bootstrapcdn	32ms	32ms	0ms
s.yimg	10ms	9ms	-1ms
static.licdn	9ms	9ms	0ms
style.aliunicorn	8ms	8ms	0ms
z-ecx.images-amazon	32ms	43ms	+11ms
Total	167ms	180ms	+13ms

Outcome observed, Span in benchmark against realworld css was marginally worse. Not worth changing.