"Exactly n" for negative n

[timsong-cpp]

copy_n and generate_n allow negative n values but then specify the complexity as "exactly n". fill_n handles this correctly.

[CaseyCarter]

D4684 [structure.specifications]/1 specifies that:

The detailed specifications of each entity defined in Clauses 7–12 follow the conventions established by
ISO/IEC 14882:2014 §17.5.1.4.

N4140 [structure.specifications] para 6 and 7:

6 If the formulation of a complexity requirement calls for a negative number of operations, the actual requirement is zero operations.

7 Complexity requirements specified in the library clauses are upper bounds, and implementations that provide better complexity guarantees satisfy the requirements.

So the specifications of copy_n and generate_n are sensible, albeit confusing. I think we can clean them up editorially. (Actually looks at copy_n) Correction, we can clean up fill_n and generate_n editorially, copy_n is just plain broken.

Proposed Resolution

Change [alg.fill] as follows:

-2 Returns: fill returns last. fill_n returns first + n for non-negative values of n and first
-  for negative values.
+2 Returns: last, where last is first + max(n, 0) for fill_n.
-3 Complexity: Exactly last - first, n, or 0 assignments, respectively.
+3 Complexity: Exactly last - first assignments.

Change [alg.generate] as follows:

 2 Returns: last, where last is first + max(n, 0) for generate_n.
-3 Complexity: Exactly last - first or n evaluations of invoke(gen) and assignments,
-  respectively.
+3 Complexity: Exactly last - first evaluations of invoke(gen) and assignments.