GCD, LCM: no endless rewriting of a lone integer; non-negative results

[tombouctou]

GCD and LCM are Flat, and their rules used the pattern ns___Integer,
which also matches a single integer among other arguments. For GCD[1/2, 3],
GCD[2.5, 5] or LCM[1/2, 3] the evaluator kept rewriting GCD[3] to 3 and
back until $IterationLimit (in a session with a time limit: until the
limit).

The rules now take all the arguments, combine the integer ones (as
before: GCD[6, 9, x] is GCD[3, x]) and leave the expression alone when
there is nothing to combine.

GCD[-4] and LCM[-4] also gave -4; they are 4.

To reproduce on master:

GCD[1/2, 3]   (* $IterationLimit::itlim, or runs until a time limit *)
GCD[2.5, 5]   (* same *)
GCD[-4]       (* -4; should be 4 *)

Tests: new cases in test/builtin/intfns/test_divlike.py (test_divlike). They fail on current master and pass with this change.

The full pytest suite and the doctests pass with this change (run together with the other fixes from the same audit; the only pytest failures, in test_main.py and test_returncode.py, are the same as on master and come from the mathics script not being on PATH in our environment).

Found in a correctness audit of Mathics3 (differential checks against SymPy/mpmath/exact computation, and calls of every builtin with edge-case arguments).

🤖 Generated with Claude Code

[rocky]

Please do the magic to add meaningful assert failures.

[tombouctou]

Done in 90a03f0: every new GCD/LCM case has a failure message saying what it checks.

[rocky]

I don't think this evaluates, but rather creates and returns an expression.

[tombouctou]

Right, it builds and returns the result rather than evaluating. Rewrote the docstring in 90a03f0 to say what comes back: an Integer when all arguments are integers, otherwise a new GCD[...]/LCM[...] with the integers combined, or None when there is nothing to combine.

[rocky]

LGTM