Add ccache for cmake caching + optimise dockerfile ordering for cache usage

[spwoodcock]

Dockerfile caching

• While debugging Ensure we use GDAL 3.11.1 in Python bindings after 24.04 upgrade #1958 it was a bit of a pain having to do lots of slow rebuilds for testing.
• Builds previously took about 2529.5s with buildkit on my Framework 16. After the PR for GDAL 3.11.1 building, it takes 2823.0s.
• The initial build after this PR took a marginally longer 3212.5s, with subsequent builds taking a much shorter 1976.9s (~46m --> 33m).
• I tested again modifying a version in requirements.txt using the ccache cache, giving 1626.3s build time - big improvement!

CI caching

• I took the opportunity to add some caching to the primary docker publish workflow.
• As a result, I upgraded most of the workflow versions, which are 5yrs outdated.
• The only breaking change I saw was with the metadata action - updated.

Note

The windows build and GPU build CI workflows also need updating to match this.
Testing with the single workflow first, then we can propagate across.

[originlake]

I'm working on something similar to optimize the docker build steps, but mostly trying to make the start-dev-env script a bit more easier to use. Currently if you delete the odmdev container, you have to rerun everything from scratch, really inconvenient. I'm trying to make it a 3 stages build and the first stage, which only installs the dev dependency can be saved for use of start-dev-env.sh. Can also publish the dev stage as an image so it can be faster for anyone wants to setup the dev environment.

@spwoodcock @smathermather I think I'll continue my work after this PR is merged, but below is my work in progress, if you can provide feedback, it would be much appreciated

originlake@2f2aba6

[spwoodcock]

Thanks @originlake!

Sorry for the slow reply with the Christmas break 😄

The changes you propose could be useful. I also planned to optimise the builds a bit, after finishing my PR for using uv to track dependencies.

Feel free to make a PR that targets this branch if you wish to extend from the ccache optimisation 🙏