ArrayFormat broken on Integer and Double (at least) data types

[bernecky]

The ArrayFormat:format() function is broken for matrix arguments. There
are two direct causes of this failure:

1. Sac2c Issue #2313 shows a broken sac2c typechecker that generates an AKV result for abs((-2)31). Since there is no way to represent 231 as an integer, the two choices I see here are either to signal an error within the typechecker, or to leave the result as AKS. The latter would, at least, defer the failure, or eliminate it if the abs() turns out to be dead code. The former approach is definitely wrong, and unpleasant. See #2313.

2. My format() code was lifted directly from the APL model, which would generate a real-number result for the above abs() case, in an interpreter. I am redesigning the format() algorithm to eliminate the use of abs(). That is work in progress...

[sbscholz]

I am afraid #2313 would require a costly change in semantics (as explained there). Therefore, the only real option is to avoid abs if that is possible....