Cohabitation of secure element interfaces is poorly defined

[gilles-peskine-arm]

There are currently two ways to define secure element drivers for PSA: the old-style dynamically registered way (enable MBEDTLS_PSA_CRYPTO_SE_C, register a driver with psa_register_se_driver(location, …)) and the new-style “unified driver interface” way (enable MBEDTLS_PSA_CRYPTO_DRIVERS, register a driver with {"type": "opaque", "location": …}).

We don't test what happens if both options are enabled. The behavior if a dynamically registered driver and a statically registered driver are registered for the same location is undefined.

We should define the behavior (presumably an error at dynamic registration time).

There should be a way to find out whether a location is in a statically registered driver, in a dynamically registered driver, or unassigned.