Report a diagnostic when a non-finite float constant is cast to an integer (#229)

[zimri-leisher]

Fixes #229.

const_convert_type caught the ValueError that int() raises for NaN but
not the OverflowError it raises for infinity, so casting an infinite constant
to any integer type escaped as an uncaught Python exception in every position
that coerces the cast (assignments, command arguments, array indices, range
bounds, exit and assert codes). Listing OverflowError alongside
ValueError gives it the same diagnostic the NaN case already gets:
For type F64: cannot convert float infinity to integer.

Red/green: the two infinity sequences crash with OverflowError: cannot convert float infinity to integer before the change and report a compile error after;
the NaN sequence guards the branch that already worked. They reach infinity by
overflowing a float multiply/add rather than through an out-of-range literal, so
they stay valid independently of #110.

🤖 Generated with Claude Code