[microTVM] Track errors on device runtime that cause resets

[areusch]

A common class of errors on embedded devices are those which cause unrecoverable interrupts (e.g. HardFault and cousins on Cortex-M). Host-driven usage of microTVM are particularly prone to these (e.g. allocating too much memory, badly implemented schedules, incorrect arguments, etc). We should add some functionality to microTVM to catch exceptions.

@mehrdadh started on this with #7671

cc @alanmacd @gromero