implement the method "is_left_modular" for lattices

[fchapoton]

extracted from #37601

📝 Checklist

• The title is concise and informative.
• The description explains in detail what this PR is about.
• I have linked a relevant issue or discussion.

[github-actions]

Documentation preview for this PR (built with commit fd7eab7; changes) is ready! 🎉
This preview will update shortly after each push to this PR.

[mantepse]

Looks good, I tested against oeis and findstat.

I just wanted to suggest that we change the signature to allow only testing semimodularity of a single element, but then I saw is_modular, so the interface is good, too.

Oh, we already have is_left_modular_element, with a slightly different implementation. It's a bit strange, it seems more than twice as fast on lattices with 9 elements:

sage: from sage.databases.findstat import _finite_lattices
sage: l = list(_finite_lattices(9))
sage: %timeit a = [L.is_left_modular([x]) for L in l for x in L]
5.25 s ± 146 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)
sage: %timeit b = [L.is_left_modular_element(x) for L in l for x in L]
1.89 s ± 39.9 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)

[fchapoton]

The method is_left_modular_element was only using the cover relations. Now, I do the same in the new method.

[mantepse]

Thank you!