feat: aggregate and summarize metrics with rust

[jonathanpwang]

closes INT-2940

Now individual benchmark-call.yml is only responsible for uploading metric json to s3 (we could use github artifact, but we're already using s3).
In benchmark.yml, we use openvm-prof to generate markdown for each benchmark from metric json and then create a summary markdown all at once.

Now that markdown creation is all in benchmark.yml, I made benchmark.yml allow workflow dispatch for ease of use (but untested).

[github-actions]

group	app.proof_time_ms	app.cycles	app.cells_used	leaf.proof_time_ms	leaf.cycles	leaf.cells_used
verify_fibair	1,626	(+634 [+0.3%]) 195,373	(+109342 [+1.4%]) 8,122,104	-	-	-
fibonacci_program	(-57 [-0.8%]) 7,337	1,500,137	51,505,102	-	-	-
regex_program	22,262	4,190,904	165,028,173	-	-	-
ecrecover_program	(-33 [-1.1%]) 2,915	290,016	15,230,037	-	-	-

Commit: 6e135a0

Benchmark Workflow