cpu/cortexm_common: make SVC call always available

[dylad]

Contribution description

This PR proposes to make SVC call always available for user and thus removes the cortexm_svc pseudomodule.
SVCall is available on all ARMvx-M architecture. This exception was used before to trigger PendSV for the sheduler. Now PendSV is triggered manually and no longer require to use SVCall.
Since the feature was provided by default for all cortexm, I think it makes sense to remove the pseudomodule and to let it hardcoded in the code.

Testing procedure

CI should be happy.

Issues/PRs references

None.

[Teufelchen1]

Looks reasonable, didn't test it

[riot-ci]

Murdock results

✔️ PASSED

4dc8a4a pkg/wamr: replace cortexm_svc FEATURE by arch_arm

Success	Failures	Total	Runtime
10177	0	10178	14m:15s

Artifacts

• Documentation preview

[dylad]

I didn't know arch_arm would also include ARM7TMDI.
I've added a fixup commit to blacklist arch_arm7 from pkg/wamr.
I'll squash if Murdock is happy about the fix.

[Teufelchen1]

Hey @dylad what do you need to move this forward?

[dylad]

I forgot about this one !
If ACK is still valid, I just need to squash this.

[Teufelchen1]

I can just give you second ACK :3

[dylad]

@kaspar030 Any objections here ?

[dylad]

Looks like Murdock failed because this PR introduces some overhead.
But cortemx_svc was already there by default so I'm quite confused with this.

[dylad]

I'm starting to reconsider this. I was wrong thinking that cortexm_svc was pulled by default already. This is NOT the case.
Indeed the feature is provided for all Cortex-M but there is no FEATURES_REQUIRED=cortexm_svc anywhere. Thus, the default mode is still the legacy one.
Switching to the new behavior will increase ROM by A LOT. Since we do not have a real use case for SVC exception yet. (except for context switching) This PR is not worthwhile.
I'm closing this.

[Teufelchen1]

That's what I get for not thinking stuff through 🥲