Allows triangular morphisms to have domain and codomain with different index sets

[hivert]

Currently triangular morphisms required to have domain and codomain with the same index sets. The patch remove this limitation by having TriangularMorphism accept a new argument inverse_on_support. It is supposed to be a function r : J |--> I returning the leading term i from the one of j = on_basis(i). This is used by: preimage and section.

I also wrote the methods co_kernel_projection, co_reduced, and _test_triangular.

Depend on #8881 and #8742, Apply patch in that order:

• ​trac_8876-triangular_morphisms_improve-fh.patch
• ​trac_8876-reviewer_patch-jb.patch

[jbandlow]

Hi Florent, I found some issues with the file as I was going through it, which I've fixed in the attached reviewer patch. (Many were probably my fault to begin with.) Everything else looks good to me. If you approve my reviewer patch and verify that all tests pass on massena without prereqs, you can set a positive review. (I will post my reviewer patch on the combinat queue immediately below yours right now.)

[hivert]

Everything was ok with the following patch applied:

trac_8881-functorial_constructions-nt.patch
trac_8742-lazy_format-fh.patch
trac_8742-lazy_format-review-nt.patch
trac_8876-triangular_morphisms_improve-fh.patch
trac_8876-reviewer_patch-jb.patch

Thanks for the review and the corrections. Your patch is good. Positive review