Request to add support for multiple threads per device

[AndMetal]

This is mostly for Vegas, but may also be useful for other cards with larger amounts of memory.

Similar to the CryptoNight miners, I think there is performance to be gained from running multiple mining threads on a single GPU. For example, running on Compute 1 & 2 of a Vega 56/64 increases hashrate by 50-75% vs running a single thread on Compute 1.

I tried simply running a second instance of lolMiner against the same device and it appears to consume double the memory (~6.6 GB on a Vega 64) but only runs on Compute 1 and after a minute or two the hashrate on both miners goes to 0 until one is stopped. So it can't currently be achieved manually that way.

I'm not sure about the architectural nuances, but this would be a huge win for AMD cards if it works (Vegas on par with 1080tis, 570s on par with 1070s, etc).

[zorzgeo]

Wote + for multiple threads!