Reduce Deserializer CPU Usage by almost 3x

[agnosticdev]

Reduce the deserializer CPU usage by almost 3x. Added a benchmark for incremental investigation.

Top of tree:

434.41 M 89.1%	52.65 M	      specialized static Deserializer<>.deserialize<>(_:claim:_:)	
298.24 M 61.1%	13.00 M	       static Deserializer<>.deserialize<>(_:_:)	
262.99 M 53.9%	12.00 M	        specialized static Deserializer<>.deserialize<>(_:_:)	
22.24 M  4.6%	22.24 M	        partial apply for closure #1 in runtest()	
83.53 M  17.1%	72.53 M	       Frame.bytes.getter	
11.00 M  2.3%	11.00 M	        Frame.startOffset.getter	
15.83 M  3.2%	15.83 M	      specialized static Deserializer<>.deserialize<>(_:_:)	
8.00 M   1.6%	8.00 M 	     specialized static Deserializer<>.deserialize<>(_:claim:_:)	

With this change:

154.37 M 69.8%	8.55 M 	      static Deserializer<>.deserialize<>(_:claim:_:)	
145.81 M 66.0%	34.19 M	       specialized static Deserializer<>.deserialize<>(_:_:)	
107.62 M 48.7%	9.77 M 	        partial apply for closure #1 in runtest()	
97.86 M  44.3%	-      	         closure #1 in runtest()	
43.31 M  19.6%	43.31 M	          specialized Deserializer<>.uint8(_:)	
24.36 M  11.0%	24.36 M	          specialized Deserializer<>.uint32NetworkByteOrder(_:)	
16.18 M  7.3%	16.18 M	          specialized Deserializer<>.uint64NetworkByteOrder(_:)	
14.00 M  6.3%	14.00 M	          specialized Deserializer<>.uint16NetworkByteOrder(_:)	
4.00 M   1.8%	4.00 M 	        specialized Deserializer<>.finalResult.getter	
25.28 M  11.4%	15.32 M	      static Deserializer<>.deserialize<>(_:claim:_:)	

[tfpauly]

Was this accidentally changed before?

[agnosticdev]

It was... by me...

[tfpauly]

Thanks for fixing =) Just confirming

[tfpauly]

Benchmark build failing with

error: 'runtest()' is only available in iOS 26 or newer
    runtest()

[agnosticdev]

Benchmark build failing with

error: 'runtest()' is only available in iOS 26 or newer
    runtest()

Addressed in 6c958f8