change logging default to be ITM instead of semihosting

[bcantrill]

No description provided.

[cbiffle]

I second Laura's concern about kernel panics becoming silent with this change, but otherwise it looks good.

[bcantrill]

At this point, this is good to go, with ITM having been verified as functioning correctly on both the STM32F407 (via STLink/v2 + Humility) and the LPC55S69 (via a cross-linked STLink/v2 + Humility).

[cbiffle]

This looks good. As a robustness thing, I feel like we ought to

• Set bit 0 of the ITM_TPR to block unprivileged (i.e. task) accesses to stimulus ports 0-7
• Move user logging to port 8.

Just because having a user insert stuff into what is otherwise the kernel (or msg trace!) log stream would be confusing.

This can be a followup if you'd like afaic.

[bcantrill]

That's a very good suggestion, but if you don't mind, I might defer that to my current work on ITM-based system call tracing, in which I would like to rationalize stimulus port usage across the entire system. In particular, we may wish logging to be done outside of a task and by a single entity: the ITM FIFO is shared by all stimulus ports, giving rise to a potential race condition whereby the FIFO can be overflowed when it is allowed to be written to by multiple tasks.