Runs-on for more actions

[blaginin]

Follow up on #20107: switch more actions to the new flow

Job	OLD	NEW	Delta
linux build test (from #20107)	3m 55s	1m 46s	-2m 09s (55% faster)
cargo test (amd64) (from #20107)	11m 34s	3m 13s	-8m 21s (72% faster)
cargo check datafusion features	11m 18s	6m 21s	-4m 57s (44% faster)
cargo examples (amd64)	9m 13s	4m 35s	-4m 38s (50% faster)
verify benchmark results (amd64)	11m 48s	4m 22s	-7m 26s (63% faster)
cargo check datafusion-substrait features	10m 20s	3m 56s	-6m 24s (62% faster)
cargo check datafusion-proto features	4m 48s	2m 25s	-2m 23s (50% faster)
cargo test datafusion-cli (amd64)	5m 42s	1m 58s	-3m 44s (65% faster)
cargo test doc (amd64)	8m 07s	3m 16s	-4m 51s (60% faster)
cargo doc	5m 10s	1m 56s	-3m 14s (63% faster)
Run sqllogictest with Postgres runner	6m 06s	2m 46s	-3m 20s (55% faster)
Run sqllogictest in Substrait round-trip mode	6m 42s	2m 38s	-4m 04s (61% faster)
clippy	6m 01s	2m 10s	-3m 51s (64% faster)
check configs.md and ***_functions.md is up-to-date	6m 54s	2m 12s	-4m 42s (68% faster)

[blaginin]

this one is quite lightweight, i switched to just 8 cpu and got the same ish time

[alamb]

Thank you for pushing this along @blaginin -- it is a really nice quality of dev improvement

[alamb]

One thing I was thinking is that someone in the future may run into this and have a hard time understanding what all the runs-on is all about

I didn't see it documented anywhere -- https://github.com/search?q=repo%3Aapache%2Fdatafusion%20runs-on&type=code

Is there any chance that you could write up some documentation in a README that explains why we are using runs-on and how to apply/unapply it?

For example, it might not be obvious without the context of this PR, to know that

  runs-on: ${{ github.repository_owner == 'apache' && format('runs-on={0},family=m8a,cpu=16,image=ubuntu24-full-x64,extras=s3-cache,disk=large,tag=datafusion', github.run_id) || 'ubuntu-latest' }}

Was the same as

    runs-on: ubuntu-latest

Except that in the apache repo it triggers the use of special runners

[blaginin]

will do!

[blaginin]

Updated docs/source/contributor-guide/index.md and referenced that in CI yml file. Do you think we should also mention that in readme?

[blaginin]

Going to merge this PR now but happy to expand docs further in the follow up in you like 🙂

[blaginin]

We just got our AWS limits increased to a very high number, so we can now switch all the remaining actions 🙂. In this PR I've switched everything big (>= 4min) except for:

• extended tests: will do in a follow-up PR
• macOS: cannot switch yet, will think about optimizations separately

I think overall this change should save us 1 hour of CI time per commit