Add benchmark tool.

[davidmorgan]

For #3800

Example output

built_value
libraries	clean/ms	no changes/ms	incremental/ms
1	20352	2872	4018
100	23670	3012	6433
250	29385	3183	12855
500	43299	4126	33312
750	70317	6590	66698
1000	110399	9757	113382

freezed
libraries	clean/ms	no changes/ms	incremental/ms
1	21249	2933	5208
100	22365	3007	6527
250	24815	3156	10736
500	37032	3498	21278
750	50651	4921	41368
1000	73173	6513	65284

json_serializable
libraries	clean/ms	no changes/ms	incremental/ms
1	21573	2840	4278
100	23040	2974	6327
250	28098	3285	12578
500	42940	4288	35061
750	69483	6561	67670
1000	115941	9178	119308

mockito
libraries	clean/ms	no changes/ms	incremental/ms
1	20752	2964	3998
100	21182	3055	5841
250	23911	3055	7754
500	29450	3335	13729
750	34840	3759	22373
1000	46205	5003	30148

[github-actions]

PR Health

Changelog Entry ✔️

Package	Changed Files

Changes to files need to be accounted for in their respective changelogs.

[davidmorgan]

Also @jensjoha who is going to start looking at build_runner things from the CFE side.

[jakemac53]

Fwiw, this kind of incremental build is probably a lot slower than a "watch" mode incremental build, because this is going to do a lot of startup work (reading asset graph, scanning and hashing all files to check for changes, etc).

Ideally, we would be able to exclude that (or separate it out).

[davidmorgan]

Yup yup, we probably want finer-grained benchmarks as we go.

For now this seems to give a clear enough signal :)

Thanks.