make tile 0 be the only tile clearing .bss

[marchioa]

At the moment all tiles clear the .bss section in crt0.S. This may cause issues when reading a vector allocated in .bss, since a slower tile might overwrite its contents when starting its routine.
This PR solves the issue by having the only tile 0 clearing the .bss.

[Dequino]

This change as-is will trigger some errors in the current test folder (for example, in test_mm_is_2).

This is mostly due to the fact that there is no initial barrier in most of the tests, which may result in the core 0 cleaning up one of the static vectors of one of the faster cores.

Two possible solutions would be:

Option 1 - Modify all tests to have an initial barrier
Naive solution as from now on all the tests would require a barrier too.

Option 2 - Add a barrier inside the crt0 itself
This can be done using a spinlock mechanism or similar.

UNFORTUNATELY atm cv32e40p does NOT support the atomic extension, so this mechanism would have to be somehow done without using atomic instructions.