Removes a few inclusion dependencies from the headers or replaces
existing ones with ones that don't indirectly include the required
headers.

This allows removing an inclusion of core/memory.h, meaning that if the
memory header is ever changed in the future, it won't result in
rebuilding the entirety of the HLE services (as the IPC headers are used
quite ubiquitously throughout the HLE service implementations).

Port citra-emu/citra#4630: "Memory: don't lock hle mutex in memory read/write"

 gl_rasterizer: Unbind textures but don't apply the gl_state 

 logging/backend: Move CreateEntry into the Impl class. Relocate local static to a class variable

web_service: Remove unnecessary inclusions

Reduces the potential amount of rebuilding necessary if any headers
change. In particular, we were including a header from the core library
when we don't even link the core library to the web_service library, so
this also gets rid of an indirect dependency.

Port citra-emu/citra#4647: "citra_qt/main: make SPEED_LIMIT_STEP static constexpr"

Moves local global state into the Impl class itself and initializes it
at the creation of the instance instead of in the function.

This makes it nicer for weakly-ordered architectures, given the
CreateEntry() class won't need to have atomic loads executed for each
individual call to the CreateEntry class.

This function is only ever used within this source file and makes it
easier to remove static state in the following change.

vk_buffer_cache: Fix clang-format

Yuzu can render 3D.

MSVC does not seem to like using constexpr values in a lambda that were declared outside of it.
Previously on MSVC build the hotkeys to inc-/decrease the speed limit were not working correctly because in the lambda the SPEED_LIMIT_STEP had garbage values.
After googling around a bit I found: https://github.com/codeplaysoftware/computecpp-sdk/issues/95 which seems to be a similar issue.
Trying the suggested fix to make the variable static constexpr also fixes the bug here.

Yuzu can now render 3D graphics to some degree.

The comment already invalidates itself: neither MMIO nor rasterizer cache belongsHLE kernel state. This mutex has a too large scope if MMIO or cache is included, which is prone to dead lock when multiple thread acquires these resource at the same time. If necessary, each MMIO component or rasterizer should have their own lock.

fuck git for ruining my day, I will learn but I will not forgive

vk_buffer_cache: Implement a buffer cache

This buffer cache is just like OpenGL's buffer cache with some minor
style changes. It uses VKStreamBuffer.

yuzu/compatdb: Remove unused lambda capture

service/audio: Provide an implementation of ExecuteAudioRendererRendering

service/audio/audren_u: Implement OpenAudioRendererAuto

This currently has the same behavior as the regular
OpenAudioRenderer API function, so we can just move the code within
OpenAudioRenderer to an internal function that both service functions
call.

This service function appears to do nothing noteworthy on the switch.
All it does at the moment is either return an error code or abort the
system. Given we obviously don't want to kill the system, we just opt
for always returning the error code.

Unbinding was required when OpenGL delete operations didn't unbind a
resource if it was bound. This is no longer needed and can be removed.

service/hid: Amend forward declaration of ServiceManager

 vk_stream_buffer: Implement a stream buffer

Improve the Accuracy of the Rasterizer Cache through a Texception Pass

gl_rasterizer/vk_memory_manager: Silence -Wreorder warnings

rasterizer_cache: mark reinterpreted surfaces and add ability to reload marked surfaces on next use.