TST: improving reliability of `oldestdeps` ?

[neutrinoceros]

hey @pllim, I noticed you had to manuallyfix oldestdeps in 2fb9a77 which, as I gathered, served as a patch against a stealth upgrade (pyparsing being a transitive dep, it's never pinned). We could instead apply the same solution I used in astropy, using uv (+ tox-uv to avoid ditching tox) to re-define oldestdeps as synonymous with a resolution strategy (lowest), so this failure mode is obliterated once and for all. I understand this repo is under-maintained, which I think is all the more reason to make its CI as stable as we can. What do you think ?

[pllim]

astroplan maintenance decision isn't my call to make. I'll defer to @eteq