refactor: convert profiling script to standard go benchmark

[kaushik-rohit]

#108

Building on previous PR #175, I updated the code to be standard Go benchmarking and addressed the comments in that PR.

Profiling run:

goos: darwin
goarch: arm64
pkg: github.com/llm-d/llm-d-kv-cache/tests/profiling/kv_cache_index
cpu: Apple M3 Pro
BenchmarkInMemory_Add-12        	     192	   6086106 ns/op
BenchmarkInMemory_Lookup-12     	       6	 178765021 ns/op
BenchmarkRedis_Add-12           	       4	 252224448 ns/op
BenchmarkRedis_Lookup-12        	       7	 143262946 ns/op
BenchmarkCostAware_Add-12       	     128	  11420663 ns/op
BenchmarkCostAware_Lookup-12    	       3	 440866472 ns/op
PASS
ok  	github.com/llm-d/llm-d-kv-cache/tests/profiling/kv_cache_index	13.403s

[yankay]

HI @kaushik-rohit

Would it be better to add a README to describe how to run it?

[kaushik-rohit]

@yankay That makes sense, I added a README

[yankay]

@yankay That makes sense, I added a README

Thanks @kaushik-rohit

It's a great change :-)
/lgtm

The test result on my PC:

(.venv) root@gpu-3090:~/oss/llm-d-kv-cache/tests/profiling/kv_cache_index# go test -bench=. -benchmem
goos: linux
goarch: amd64
pkg: github.com/llm-d/llm-d-kv-cache/tests/profiling/kv_cache_index
cpu: AMD Ryzen 9 3900 12-Core Processor
BenchmarkInMemory_Add-24                      45          23630930 ns/op        13737528 B/op     130167 allocs/op
BenchmarkInMemory_Lookup-24                    2         615924412 ns/op        2564472368 B/op   131676 allocs/op
BenchmarkRedis_Add-24                          3         375040929 ns/op        21154280 B/op     556949 allocs/op
BenchmarkRedis_Lookup-24                       7         185071392 ns/op         9064401 B/op     260133 allocs/op
BenchmarkCostAware_Add-24                     26          43422847 ns/op        11363554 B/op     132215 allocs/op
BenchmarkCostAware_Lookup-24                   1        1131001154 ns/op        2564395856 B/op   160103 allocs/op
PASS

[yankay]

HI @sagearc @hyeongyun0916 @liu-cong @vMaroon

Would you please help to review it :-)

[vMaroon]

Thanks @kaushik-rohit, @yankay, apologies for delay.

/lgtm
/approve