Conv convention

[t6s]

Motivation for this change

This PR changes the convention of the convex combination operation in convex.v from
a <| t |> b = (1-t)a + tb to ta + (1-t)b, to make this compatible with infotheo (https://gihtub.com/affeldt-aist/infotheo).

There has been a discussion about these two options, and the former
has been chosen for its compatibility with the homotopical notion of paths,
while the latter is algebraically natural.

After a discussion in a analysis-dev meeting, a conclusion was made to change it to the latter.
Also, there is now a line_path in set_interval.v which essentially implements the former convention
and has an appropriately geometric name.

Therefore I think it is now reasonable to do this PR.

@CohenCyril @affeldt-aist

Checklist

• added corresponding entries in CHANGELOG_UNRELEASED.md

• added corresponding documentation in the headers

Reference: How to document

Merge policy

As a rule of thumb:

• PRs with several commits that make sense individually and that
  all compile are preferentially merged into master.
• PRs with disorganized commits are very likely to be squash-rebased.

Reminder to reviewers

• Read this Checklist
• Put a milestone if possible
• Check labels

[affeldt-aist]

Great! This will unlock the port of convex spaces from InfoTheo. Regarding the PR in itself, the technical lemmas about the parameters of the quasi-associativity should be better named and documented. However, I think that it will be more efficient for us to do that along the port of more properties of convex spaces. So I am ok to merge with the proviso that we agree to carefully review that part of the library again in the near future.

[t6s]

I am afraid of naming the submodule QuasiAssoc, observing there are other (likely to be many other) variations of "quasi-associativity" in algebra. E.g. https://academic.oup.com/ptep/article/2014/1/013A03/1552583

[affeldt-aist]

Let's be more precise and even hide the module name behind a definition in lowercase akin to associative, etc.