kernel/process: Set ideal core from metadata

savedata_factory: Add CacheStorage and delete TemporaryStorage on boot

vm_manager: Make vma_map private

gl_shader_decompiler: TLDS/TLD4/TLD4S Reworked reflecting the source registers, bugs fixed and modularize.

service/ldr: Amend layouts of NRO and NRR headers

hle/service, hle/sm: Minor cleanup

Backport review comment from citra-emu/citra#4418

loaders: Make GetFileType() a const qualified member function

gl_shader_decompiler: TLDS/TLD4/TLD4S Reworked reflecting the source registers, bugs fixed and modularize.

Original reason:
As Windows multi-byte character codec is unspecified while we always assume std::string uses UTF-8 in our code base, this can output gibberish when the string contains non-ASCII characters. ::OutputDebugStringW combined with Common::UTF8ToUTF16W is preferred here.

Mimics hardware behavior.

configuration/config: Use an intermediary variable for accessing players

system_archive: Implement open source NgWord2

Given memory should always be expected to be valid during normal
execution, this should be a debug assertion, rather than a check in
regular builds.

This was only ever public so that code could check whether or not a
handle was valid or not. Instead of exposing the object directly and
allowing external code to potentially mess with the map contents, we
just provide a member function that allows checking whether or not a
handle is valid.

This makes all member variables of the VMManager class private except
for the page table.

 kernel/svc: Correct behavior of svcResetSignal() 

gl_rasterizer: Implement a framebuffer cache

gl_shader_decompiler: Implement TEXS.F16

Combines the two into one, shortening the amount of code here.

Only one usage of the specified objects made use of the lack of
namespacing. Given the low usage, we can just remove these.

These auto-deduce the result based off its arguments, so there's no need
to do that work for the compiler, plus, the function return value itself
already indicates what we're returning.

Gets rid of the need to keep the variables separate from their actual
initialization spots.

 system_archive: Use a regular function pointer instead of std::function for the file-scope system archive array

service/ldr: Deduplicate instruction cache clearing code in LoadNro()

Perf: Call shrink_to_fit after page-table vector resizing to actually reduce vector capacity

Similarly, here we can avoid doing unnecessary work twice by retrieving
the file type only once and comparing it against relevant operands,
avoiding potential unnecessary object construction/destruction.

While GetFileType() is indeed a getter function, that doesn't mean it's
a trivial function, given some case require reading from the data or
constructing other objects in the background. Instead, only do necessary
work once.

No implementations actually modify instance state (and it would be
questionable to do that in the first place given the name), so we can
make this a const member function.

Call shrink_to_fit after page-table vector resizing to cause crt to actually lower vector capacity. For 36-bit titles saves 800MB of commit.

A very trivial change. If metadata is available, the process should use
it to retrieve the desired core for the process to run on.

Avoids typing the same long accessor just to retrieve player attributes.

We've already given the constant to the vector itself, so we don't need
to re-hardcode it in the array.

system_archive: Use a regular function pointer instead of std::function for file-scope system archive array

This allows the array to be constexpr. std::function is also allowed to
allocate memory, which makes its constructor non-trivial, we definitely
don't want to have all of these execute at runtime, taking up time
before the application can actually load.

We don't need to specify all of the ARM interfaces explicitly.

The first word is just a padding byte, it's not an actual entry
instruction. Also renames the rest of the entries according to
SwitchBrew.

The padding after the magic signature value should be 12 bytes rather
than 28 bytes. The other 16 should be placed after the title ID pattern.