Refactor NVML + Health bits

[Spudz76]

Allow unavailable items to disappear

• various GPUs are missing features; detect via checking the return value for nonsuccess, then tarnish the health bucket by setting it to literal maximum as a token for unsupported/broken (and don't continue to query them) Tested on a laptop which does not report fans due to integrated unified cooling system, and does not have PowerUsage reporting because it's Fermi, etc. All it shows is temperature, now, perfect.
• refactor/simplify worker health output to build the string incrementally (more easily handle all the possibly missing features) and dump at the end

Make thermal display color thresholds configurable

• compile time via new src/defaults.h
• runtime via temp_low and temp_high within advanced thread config

and random patches to anything that stuck out while I was in the area(s) half of which my editor just does for me anyways (actually harder to NOT do them)

[xmrig]

I will merge it today evening or tomorrow, need some time for it, because I support fully functional command line interface, adding new options not that simple.
Thank you.

[Spudz76]

I can add that too, I just didn't. Thought it would be safer to have it harder to edit?
Eventually adding high temp exit trigger etc and that's like a thermostat it should be in a plastic box so the kiddies can't tweak it by accident.

[Spudz76]

and mainly 45-65 is not my favorite range so I made it compile-time feature first

really just remove the configuration part, I don't want it there

hardcoded only.

[Spudz76]

I only added it into the threads tree since that's the only place to store them per-GPU
and if you have mixed GPUs
some old crap likes to be 88c but I would certainly get tired of red letters quick since they are probably over 65 at the first job assignment
Heck my GTX970 is 72c and it's underclocked

I just never saw those colors change and wondered if the range was right, if they would... :)

[Spudz76]

The best part is I run everything through meta-miner so it strips off any color anyways so I literally don't see it

[xmrig]

I get multiple issues:

1. MAXUINT32 is Windows specific, better use std::numeric_limits<uint32_t>::max() or external flags for available features, values like 4294967295 leaks to API.
2. unsigned __int64 -> uint64_t.
3. Some formatting issues, like 94W54C.
4. dynamic_cast -> static_cast.

I decided reject this PR in current state, but I agree with you, temperatures should be configurable, probably better names for options is temp-warning and temp-danger or something like that.
This feature will be added to v2.15.
Thank you.

[xmrig]

Also:

/home/.../xmrig-nvidia/src/workers/Workers.cpp: In function ‘const string _spf(const char*, ...)’:
/home/.../xmrig-nvidia/src/workers/Workers.cpp:135:20: error: array must be initialized with a brace-enclosed initializer
     va_list args = nullptr, copy = nullptr;
                    ^~~~~~~

[Spudz76]

Applied much of the cool stuff onto #257 and then some
with none of the complainy bits (I mostly agree)
probably will fixup some of those bits

definitely need to test on Linux too or I'm going to keep bashing into silly problems
but I have no Linux with nvidia handy at the moment

Could have sworn MAXUINT32 was standard back to like regular C even (also swore I've used it on Linux). Or at least as soon as standard libraries had limits.h? idk, agree though the C++ standard method for referring to it is better (idk C++ rly)