Matching minor refactorings

[gasche]

In parallel to #13121, here is another small PR in the review stack to fix The Pattern-Matching Bug.

(If you wonder about the state of this review stack, that is, when this damn thing is going to be fixed or at least you won't get emails about it anymore: I maintain an overview comment that describes the current best-known state. Currently I have 4 PRs to propose after this one, and the first 2 suffice to fix the bug.)

The present PR contains minor refactorings that should obviously not change the behavior of the pattern-matching compiler.

• the first commit fixes a naming mistake I made in #13084, where I introduced the name head to refer to "the head argument of an argument list" (once it is in "simple" form), without realizing that it conflicts with the pre-existing usage of head to refer to "the head of the pattern" (also called "discriminant", and a pretty important notion in division of matrices). The commit renames head::rest into first::rest again, and the type pure_head becomes pure_arg.
• the second commit turns the tuple lambda * let_binding into a record with named fields, to prepare for the addition of a third field in an upcoming PR

[gasche]

(cc the usual suspects: @ncik-roberts , @trefis, @lthls, maybe @Octachron and @voodoos if they want to; note that #13121 is also waiting for a reviewer, but reviewing that other PR involves slightly more domain knowledge )

[gasche]

Thanks! Would you by chance be interested in approving-on-behalf for #13121 as well?