stop: grow stacks for async tasks

[tbg]

Surprisingly, this seems to have helped with alloc/op, but not with CPU time, at least in the microbenchmark.

name                                   old time/op    new time/op    delta
Sysbench/SQL/3node/oltp_read_write-24    11.6ms ± 1%    11.6ms ± 2%    ~     (p=0.201 n=14+15)

name                                   old alloc/op   new alloc/op   delta
Sysbench/SQL/3node/oltp_read_write-24    2.20MB ± 3%    2.16MB ± 1%  -1.88%  (p=0.047 n=15+12)

name                                   old allocs/op  new allocs/op  delta
Sysbench/SQL/3node/oltp_read_write-24     11.0k ± 2%     10.8k ± 0%    ~     (p=0.069 n=15+12)

End-to-end testing via the sysbench-settings roachtest (10 runs) shows a significant improvement in qps¹:

	max	mean	median	% change max	% change mean	% change median
old	23673.17	22775.40	22684.17
new	23955.21	23154.03	23256.25	1.19%	1.66%	2.52%

Closes #130663.

Epic: CRDB-43584

Footnotes

1. https://docs.google.com/spreadsheets/d/1QUBZHllhk5CtDfcsJqc5eOIqUO4jVdhoYXMWj4CRQaQ/edit?gid=0#gid=0 ↩

[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 

[tbg]

Before this PR, sysbench oltp_read_write on 3x8vcpu:

This was running on an older SHA, so the comparison isn't scientific, but I anecdotally see around 2-3% even on recent SHAs.

With this PR:

after.pb.gz

It's interesting that there are still a number of callers that need to grow the stack in addition to growstack.Grow.

[tbg]

I ran oltp_read_write 5x on #137927, but it's so choppy:

base: 21778 22081 20964 19356 17538
with: 20764 20481 18452 20808 21295

hoping that #137930 significantly reduces the choppiness, and will revisit in the new year.

[tbg]

10x sysbench-settings/oltp_read_write on this branch:

Details

$ rg --no-filename 'queries:' | grep -Eo '[0-9]+\.[0-9][0-9]' | sort -rn
23955.21
23590.72
23451.91
23312.14
23299.67
23212.83
23201.54
22806.27
22592.12
22117.93

Whereas 10x at the preceding commit:

Details

$ rg --no-filename 'queries:' | grep -Eo '[0-9]+\.[0-9][0-9]' | sort -rn
23673.17
23378.29
23160.06
22769.08
22685.97
22682.37
22621.54
22563.36
22350.77
21869.38