psa_interruptible_set_max_ops() function

The time taken to execute a single op is implementation specific and depends on software, hardware, the algorithm, key type and curve chosen. Even within a single operation, successive ops can take differing amounts of time. The only guarantee is that lower values for \p max_ops means functions will block for a lesser maximum amount of time. The functions \c psa_sign_interruptible_get_num_ops() and \c psa_verify_interruptible_get_num_ops() are provided to help with tuning this value. This value defaults to #PSA_INTERRUPTIBLE_MAX_OPS_UNLIMITED, which means the whole operation will be done in one go, regardless of the number of ops required. If more ops are needed to complete a computation, #PSA_OPERATION_INCOMPLETE will be returned by the function performing the computation. It is then the caller's responsibility to either call again with the same operation context until it returns 0 or an error code; or to call the relevant abort function if the answer is no longer required. The interpretation of \p max_ops is also implementation defined. On a hard real time system, this can indicate a hard deadline, as a real-time system needs a guarantee of not spending more than X time, however care must be taken in such an implementation to avoid the situation whereby calls just return, not being able to do any actual work within the allotted time. On a non-real-time system, the implementation can be more relaxed, but again whether this number should be interpreted as as hard or soft limit or even whether a less than or equals as regards to ops executed in a single call is implementation defined. For keys in local storage when no accelerator driver applies, please see also the documentation for \c mbedtls_ecp_set_max_ops(), which is the internal implementation in these cases. This only applies to functions whose documentation mentions they may return #PSA_OPERATION_INCOMPLETE.