Respect the direction set in a template stream

[bifurcation]

This PR attempts to fix the logical error noted in #617, that the directions set on a template stream are not checked when the template stream is used to create new (directed) streams. It simply adds checks in srtp_protect_mki and srtp_unprotect_mki that verify that the template stream's direction is compatible with the context, either the same direction or dir_srtp_unknown. Marking as WIP because even if we agree on approach, we probably need to cover the analgous RTCP methods.

Note that there is compatibility risk in merging this PR: If there are consumers that are relying the ambiguity noted here, then they will fail.

[pabuhler]

This code apears to be 17 years old and kind of unchanged, I agree that it should be fixed, but it might break things for exisitng clients if they have been doing it wrong. Maybe this is something for 3.0 ?
The update_xxx() functions also need to preserver the directions flags.

[pabuhler]

@bifurcation , I took your change and applied some more direction checking, can you do a review ?

Summary of changes:

Tightens SRTP stream direction handling so sender and receiver streams cannot be used for the wrong operation.

It adds a new srtp_err_status_direction_mismatch status for cases where an existing stream is used with the opposite direction, and applies that check consistently across RTP and RTCP protect/unprotect paths. Wildcard template streams now also reject incompatible inbound/outbound operations before creating a stream.

The change also preserves stream direction across srtp_update(), including both specific streams and wildcard template streams, and rejects template updates that would flip an established template direction.