Fix guard equality with short-circuit operators

[gldubc]

Closes #15253.

This fixes guard evaluation for andalso and orelse when they appear on the right side of a comparison. The issue happens after analysis, while of_guard re-evaluates the guard call.

The fix only uses the pattern-specific guard handling when pattern_info is available. Otherwise it dispatches through Apply.remote (as of_remote does with other cases such as the comparisons @comp_op), with signatures for andalso and orelse added there to preserve type precision.

[josevalim]

Unfortunately andalso and orelse mean boolean, term, so we need to adapt it accordingly. I will take it from here, thank you!!!!