OpenCL benchmark chooser

[ohhmm]

In SKRYPT, CPU may have more compute units then GPU:

Tesla K80:
        version: OpenCL 3.0 CUDA
        compute_units: 13
        max_work_item_dimensions: 3
        address_bits: 64
        clock_frequency: 823
        max_work_group_size: 1024

Tesla K80:
        version: OpenCL 3.0 CUDA
        compute_units: 13
        max_work_item_dimensions: 3
        address_bits: 64
        clock_frequency: 823
        max_work_group_size: 1024

NVIDIA GeForce GT 710:
        version: OpenCL 3.0 CUDA
        compute_units: 1
        max_work_item_dimensions: 3
        address_bits: 64
        clock_frequency: 954
        max_work_group_size: 1024

AMD Ryzen Threadripper 3990X 64-Core Processor :
        version: OpenCL 3.0 (Build 0)
        compute_units: 128
        max_work_item_dimensions: 3
        address_bits: 64
        clock_frequency: 0
        max_work_group_size: 8192

Intel(R) FPGA Emulation Device:
        version: OpenCL 1.2
        compute_units: 128
        max_work_item_dimensions: 3
        address_bits: 64
        clock_frequency: 0
        max_work_group_size: 67108864

Selected AMD Ryzen Threadripper 3990X 64-Core Processor
        max_work_group_size: 8192

SKRYPT RULEZ:
Accepts three forms of input :
- equations with any variables are automatically equaled to zero
- single variable name to show its value if known
- expression with question mark to check if its true

Benchmarking would make actual info on proper preference.