inconsistent synchronization api

[DenisBiryukov91]

Describe the release item

There are several inconsistencies with synchronization api (z_task..., z_convar..., z_mutex...).

• All of the types they operate upon are technically non-copyable; i.e. they can only be aliased or moved - should we enforce it more clearly using owned_ semantics ?
• z(p)_task_free accepts **z_task_t; is double-pointer really necessary here, given that other _free functions accept a simple pointer, there is no enforcement to allocate z_task_t on heap, and other z(p)_task... functions accept a simple pointer
• the behavior of z(p)_task_free does not seem to be consistent across platforms: in most cases z_task_t pointer is just freed - which likely corresponds to the thread detach if it was not join before, while in freertos and flipper - some related system resources are freed.
• Is z_task_free really needed at all, given that we can simply put all related logic in z(p)_task_join or zp_task_cancel ?

[DenisBiryukov91]

resolved by #484 and #489.