performance improvelemts & cleanup

[zgabi]

Output parsing is also chagned to use the tensor's Buffer instead of the indexer. +Some cleanup.

Performance test (with the already existing Yolov7netPerformance project:

Before:	Method	Mean	Error	StdDev	Median	Gen 0	Gen 1	Gen 2	Allocated
Yolov7	20.67 ms	0.230 ms	0.204 ms	20.64 ms	500.0000	500.0000	500.0000	5 MB
Yolov8	40.55 ms	2.130 ms	6.282 ms	42.57 ms	9000.0000	2000.0000	1000.0000	103 MB
Yolov5	20.58 ms	0.388 ms	0.399 ms	20.51 ms	750.0000	562.5000	562.5000	7 MB

After:	Method	Mean	Error	StdDev	Gen 0	Gen 1	Gen 2	Allocated
Yolov7	20.23 ms	0.401 ms	0.492 ms	500.0000	500.0000	500.0000	5 MB
Yolov8	16.67 ms	0.333 ms	0.342 ms	500.0000	500.0000	500.0000	5 MB
Yolov8Numpy	89.37 ms	3.732 ms	10.829 ms	9000.0000	1000.0000	1000.0000	110 MB
Yolov5	15.93 ms	0.217 ms	0.203 ms	500.0000	500.0000	500.0000	5 MB

[iwaitu]

great job