For some reason, currently unknown, FFmpeg logging has to be set to
something in order for HLS to not fail on Windows, even if it's just an
empty function.

(Author comment: I don't feel comfortable not knowing why, my theory is
that FFmpeg has an issue with this internally with HLS on Windows
related to logging or how it's trying to log, I have absolutely no idea
what's going on and if someone can enlighten me or debug FFmpeg itself
it'd be real nice to know why.)

Adds a cmake variable (DEBUG_FFMPEG_MUX) which enables FFmpeg debug
output in the ffmpeg-mux subprocess, and if on Windows, shows the
console window of the ffmpeg-mux subprocess so the current output can be
seen.

Fix #3514 on MacOS and Ubuntu.

For game capture, hook DXGI release function to release D3D objects if
the related swap chain is also being destroyed.

An added bonus is that the game capture hook will handle swap chain
recreation for applications that don't use ResizeBuffers.

Add HLS streaming output

Add and register an obs_output_info struct called
ffmpeg_hls_muxer. It uses ffmpeg's HLS muxer to stream output.

Add threading and buffer to reduce skipped frames.
Also add frame drop logic when there are too many packets in the
buffer for congestion control.

(Jim) Allows the ability to use stream keys with ffmpeg-mux -- this will
enable the ability to use it with certain services (such as YouTube)
which allow different protocols (such as HLS)

(Jim) Allows the ability to specify mux settings directly on the
output's structure itself rather than just via settings. This is useful
for network-based muxer outputs (such as HLS).

(Jim) printable_file allows the ability to keep a separate string for
logging which will not contain things like stream keys when used with
outputs such as HLS.

Allows the ability to get the stream key for YouTube regardless of
whether using the normal RTMP or HLS versions of YouTube.

(Jim) Allows the ability to get a link from the service's settings about
a specific service selection the user chooses and display it as a "More
Info" button that the user can click to find out more information about
that particular service.

(Jim) Allows the ability to use it for more output types.

The frontend API event OBS_FRONTEND_EVENT_REPLAY_BUFFER_SAVED was not
being triggered when hotkeys were used because it was only set to
trigger when the replay buffer button was clicked, not when a hotkey was
used. This fixes it so that the frontend event sends a signal to the
frontend so that hotkeys will actually work.

New events must go at the bottom of the list, not at the top, otherwise
ABI will be broken.

* obs-frontend-api: add the event of saving replay buffer

Add OBS_FRONTEND_EVENT_REPLAY_BUFFER_SAVED as given by RFC 33

* UI: Emit the replay buffer saved event to the api

Send the OBS_FRONTEND_EVENT_REPLAY_BUFFER_SAVED to api (as in rfc33)

* docs/sphinx: Add replay buffer saved event

Documentation provided for OBS_FRONTEND_EVENT_REPLAY_BUFFER_SAVED (RFC33)

Adds the virtualcam GUID to win builds on GH actions and azure.
Co-authored-by: NPatTheMav <patthemav+github@gmail.com>

Docs: Fix Frontend Finish Loading event

Fixes obsproject/obs-studio#3465

When deleting a transition, it would sometimes select the one that
replaces it, which could be the separator.  This would cause undefined
behavior.  This fixes it so that it selects either the one that replaces
it (if a valid one replaces it), or the transition at the bottom of the
known list and not the separator.

Apparently using LoadLibrary on a full path doesn't work -- you need to
use SetDllDirectory before loading a library.

The path of coreaudio dll has changed in current versions of iTunes on
windows.  The patch fixes the issue while keeping compatibility with
previous path.

Fixes obsproject/obs-studio#3503

Refactors the windows import library section so that it uses a list, and
remove a slightly unnecessary macro in release_lib

This reverts commit 01fe9c99.

EVGA/magewell devices seem to use the default system drivers rather than
custom drivers, which causes their audio to become decoupled and treated
as completely separate devices rather than as an audio pin on the video
device. Basically, this would cause those devices to not have audio by
default, and force the user to have to manually select the audio, which
is bad user experience.

We already had a workaround for this with elgato devices, so expand that
code to become a whitelist of devices, and include EVGA/magewell
devices.

We were performing a custom cursor draw for WGC, which we attempted
because we thought we could bring back hardware cursor responsiveness,
but MS says that it won't. Since the custom draw was bugged to cause
black screens anyway, just use the native cursor draw for now.

Rivatuner Statistics Server calls vkGetInstanceProcAddr for
vkCreateDevice with a NULL instance. According to spec, this should
return NULL, but RTSS will dereference this and crash.

This change returns function pointers for every intercepted function
regardless of instance, which is enough to keep RTSS alive.