Improvements to type marshaling

[manodasanW]

• Added benchmarks for type marshaling
• Added a cache for type name generation and made use of an existing cache for the reverse lookup
• Given the Type struct has a single field that makes it non blittable which is a string, made it a ref struct and made use of the MarshalString pinnable changes to pin the string when the ref struct for this is pinned allowing for us not need to use GC handles to do the same.
• Fixed Marshaler to special case type as there were multiple functions with the same name now which made those lookups fail.

Master

Method	Mean	Error	StdDev	Gen 0	Allocated
GetWinRTType	1,237.2 ns	8.85 ns	7.39 ns	0.0572	240 B
SetWinRTType	531.9 ns	2.68 ns	2.51 ns	0.0935	392 B
SetPrimitiveType	310.9 ns	2.41 ns	1.88 ns	0.0401	168 B
SetNonWinRTType	377.4 ns	3.57 ns	2.79 ns	0.0572	240 B
GetExistingWinRTType	1,221.3 ns	8.08 ns	7.56 ns	0.0572	240 B

PR

Method	Mean	Error	StdDev	Gen 0	Allocated
GetWinRTType	93.29 ns	1.552 ns	1.451 ns	0.0286	120 B
SetWinRTType	156.65 ns	1.852 ns	1.641 ns	-	-
SetPrimitiveType	129.65 ns	2.387 ns	2.233 ns	-	-
SetNonWinRTType	158.73 ns	3.163 ns	4.435 ns	-	-
GetExistingWinRTType	94.83 ns	1.730 ns	1.533 ns	0.0286	120 B

.NET core 3.1 (previous built-in support)

Method	Mean	Error	StdDev	Gen 0	Allocated
GetWinRTType	3,829.2 ns	16.27 ns	12.70 ns	0.0229	120 B
SetWinRTType	869.7 ns	7.67 ns	5.99 ns	0.0286	120 B
SetPrimitiveType	320.0 ns	2.57 ns	2.28 ns	0.0076	32 B
SetNonWinRTType	226.6 ns	0.84 ns	0.74 ns	0.0496	208 B
GetExistingWinRTType	3,858.8 ns	30.49 ns	23.80 ns	0.0229	120 B