Generalize integration_by_parts

[IshiguroYoshihiro]

Motivation for this change

Add a part of generalized version of integration by parts for unbound intervals `[a, +oo[.
This PR is over #1656 and #1662.
I don't sure that I should add 4 more lemmas and complete all variation of integration by parts for unbound intervals.
I think that 8 similar lemmas would be a mess. So, I'd like to somehow make them up into one or two lemmas, but I don't have an idea.

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

[Tragicus]

Are not all 4 versions the same, the only difference being whether the minus is inside or outside the integral?

[IshiguroYoshihiro]

Are not all 4 versions the same, the only difference being whether the minus is inside or outside the integral?

If you said that these lemmas can be proved with one of them, I think it is no.
As you can see, the version of le0_ge0 and le0_le0 can be proved with ge0_ge0 or ge0_le0, but ge0_le0 can not proved from ge0_ge0.
Because that integration by substitusion using a transformation F -> F \o -%R change the unbounded side of the interval from `[a, +oo[ to `[-oo, - a[, I may have to set 4 more variations of this lemma for `[-oo, -a[.
It will makes messy for the use to set 8 variation of the lemma. So I will consider some good way.