improve performance of RegistrationKey

[bollhals]

PR 1 extracted out of #40

Focused on the RegistrationKey. 2 main changes:

• Implement IEquateable to avoid boxing in equality / hashcode methods
• Cache the typegroup to improve the hashcode performance

Performance measurements:

From ResolveFromType, same performance, but less allocation

Method	Mean	Error	StdDev	Min	Max	Median	Gen 0	Gen 1	Gen 2	Allocated
Master	220.3 ns	0.63 ns	0.56 ns	219.5 ns	221.3 ns	220.4 ns	0.0730	-	-	344 B
Current	221.8 ns	2.19 ns	1.94 ns	220.3 ns	226.7 ns	220.9 ns	0.0577	-	-	272 B

From ResolveFromGenericType, faster performance, less allocations

Method	Mean	Error	StdDev	Min	Max	Median	Gen 0	Gen 1	Gen 2	Allocated
Current	676.1 ns	5.69 ns	5.32 ns	669.4 ns	685.6 ns	674.6 ns	0.0916	-	-	432 B
Master	877.2 ns	5.14 ns	4.81 ns	869.9 ns	884.2 ns	877.6 ns	0.1068	-	-	504 B
v1.4	829.0 ns	4.72 ns	4.42 ns	822.7 ns	838.0 ns	827.9 ns	0.0639	-	-	304 B
v1.BoDi_Concurrent_Dictionary_And_Lazy	1,026.5 ns	4.18 ns	3.70 ns	1,020.5 ns	1,032.4 ns	1,025.6 ns	0.1564	-	-	736 B