bench: ignore cluster startup and shutdown

[yuzefovich]

This commit adjusts our benchmark harness to ignore the overhead of starting up and shutting down the test cluster. This allows us to avoid spurious "regressions" like that we've seen due to #117116.

Epic: None

Release note: None

[blathers-crl]

It looks like your PR touches production code but doesn't add or edit any test code. Did you consider adding tests to your PR?

_{🦉 Hoot! I am a Blathers, a bot for CockroachDB. My owner is dev-inf.}

[cockroach-teamcity]

This change is 

[yuzefovich]

The impact of #117116 (which made the cluster shutdown much slower) is mentioned here. I picked one benchmark as a sanity check for this change, and it has the desired effect:

name                                          old time/op    new time/op    delta
WideTableIgnoreColumns/MultinodeCockroach-24     1.75s ± 4%     0.00s ± 1%  -99.79%  (p=0.000 n=10+9)

name                                          old alloc/op   new alloc/op   delta
WideTableIgnoreColumns/MultinodeCockroach-24    37.2MB ± 6%     2.3MB ± 1%  -93.78%  (p=0.000 n=9+10)

name                                          old allocs/op  new allocs/op  delta
WideTableIgnoreColumns/MultinodeCockroach-24      107k ± 1%        1k ± 1%  -98.73%  (p=0.000 n=8+9)

[mgartner]

LGTM

[yuzefovich]

TFTR!

bors r+

[craig]

Build succeeded:

• Bazel Essential CI (Cockroach)