Prioritise building more recent packages

[tmcgilchrist]

Starting a discussion here about what should happen with building older version of packages. Currently we attempt to build everything in opam-repository which includes a number of older packages that are of questionable usefulness.

For example:

• older Tezos releases
• older JaneStreet base/core libraries

Prioritise building the more recent versions of packages.
What are the OCaml versions that we support package documentation for?
Can we keep the last successful documentation build for older releases and not rebuild them?

ocaml/opam-repository#23649

[sabine]

After some point, when odoc's lower bound rises, we will most likely want to keep around the old successful documentation build for packages that will no longer build and will have to support the old data format from thereon (or convert the old data to the new format if there's changes on voodoo's output data format). 🤔 While odoc is still in active development, I think that we may want to get the new features also on older documentation that people still use. We currently don't have metrics on what people use. This seems like something we want to collect.

We should determine the exact lower bound on ocaml imposed by voodoo and odoc. The voodoo CI only tests voodoo explicitly down until 4.08, but the real lower bound is earlier than 4.08.

Older Tezos releases seem mostly useless, and I am curious why there are so many Tezos packages on opam in the first place.

I think this discussion is a great one to be had in public, similar to https://discuss.ocaml.org/t/raising-the-minimum-tested-version-of-ocaml-to-4-05-or-4-08-from-4-02/12464.