Camera rework and small fixes

[Meziu]

Big (and messy) PR incoming!

Changes related to the cam service:

• Rework of many cam functions to fix weird behaviour or crashes and to streamline the service. Still WIP (Trimming still needs some research and the image calibration configs don't seem to work).
• Removed the need for ALL redundant "image size" requests in function arguments.
• Support for BothOutwardCam using a specific take_picture implementation.
• The camera-image example now takes 3D images (and is abnormally slow thanks to CPU rendering).
• The camera-image example doesn't need all the specific const values to work properly.

Some other minor changes that were made in the mean time:

• Renamed white/blacklist functions to allow/blocklist.
• Safely wrapped Console::set_window.
• Properly documented GSPGPU events.
• Added missing functionality for the Hid module, such as volume slider and accelerometer/gyroscope. After some more research (done for the changes) I noticed the service requires to be exclusively usable from one instance, using ServiceReference, as it modifies global states which are normally un-trackable.
• Changed how ServiceReference checks for service usage. Poison handling is now a feature!
• Fixed some panic! cases that are now handled as new Error types. (also renamed NdspError to just Error following the module namespaces changes)
• Added warnings and changed functionality to avoid dangerous unsafe cases when using Gfx on the VRAM.
• Translated a couple of calls to static inlines after the merge of Support static inline functions in ctru-sys #133.

P.S.
I'm sorry for mixing up the changes to the cam service and those related to TODO or other small fixes. I had initially envisioned this PR as a "todo fix" PR but as I worked on it I noticed that cam was going to need a lot of work.

[Meziu]

It seems like nobody on github ever used stereo calibration before (not even devkitPro's own examples include it) nor image quality calibration. In my tests it seems like CAMU_SetStereoCameraCalibrationData has no effect, so I'm wondering how these functions are supposed to work. The same goes for trimming, of which I have understood only that it accepts images of the "standard" supported resolutions (like VGA and TOP_LCD) but that I still can't grasp in its entirety (does it accept all resolutions with an area multiple of 256?).

@Steveice10 Sorry for the tag, but since you've written the cam:u implementation for libctru, and have a lot of experience with FBI and cameras in general, would you mind sharing some of your knowledge with us? I've spent a couple of weeks reverse-engineering the cam service, but your insight would be greatly appreciated.

[Meziu]

Thank you so much for the information! I didn't expect your quick response 😄.

* You need to use `cam:s` (I think `cam:c` also works?) for privileged access. All 3 cam services share the same handler but have state checks on what service needs to be used to access certain commands.

Interesting. This effectively means I'd need to re-implement all IPC calls to use my own cam:s handle, correct? In libctru everything uses cam:u, so that probably won't work. Also, would that change bring any trade-offs that you know of, or is the handle to use only up to the specific needs?

2. For trimming I don't think I ever played with it thoroughly enough beyond standard dimensions to know what exactly it will and won't accept.

No problem, I've already been experimenting with it enough. I should be able to find some conclusive evidence in no time, i hope...

I guess I should try to merge this PR more or less as-is and work on re-implementing everything over cam:s if possible later, since it would be way too disruptive right now.

[ian-h-chamberlain]

I didn't have a chance to deeply read the new camera code yet but mostly LGTM! Couple minor comments but otherwise all seems good

[Meziu]

This PR is finally ready to merge.