Performance optimization for SvgPointCollectionConverter parser

[wieslawsoltes]

Reference Issue

Split from #786

What does this implement/fix? Explain your changes.

Performance optimizations for SvgPointCollectionConverter parser

dotnet run -c Release -f netcoreapp3.1 -- -f '*SvgPointCollectionConverter_*'

• with CoordinateParser

Method	Mean	Error	StdDev	Op/s	Rank	Gen 0	Gen 1	Gen 2	Allocated
SvgPointCollectionConverter_Parse	495.0 ns	5.43 ns	5.08 ns	2,020,232.5	1	0.0067	-	-	136 B

• with StringSplitEnumerator

Method	Mean	Error	StdDev	Op/s	Rank	Gen 0	Gen 1	Gen 2	Allocated
SvgPointCollectionConverter_Parse	460.4 ns	3.17 ns	2.81 ns	2,171,837.9	1	0.0072	-	-	136 B

Any other comments?

[inforithmics]

I think this can now be merged because. The Roslyn Compiler can soon do switches on ReadonlySpans see issue. https://github.com/dotnet/csharplang/issues/1881