Occassionally the auto adjust doesn't find enough memory and the
intensity is detected too low and aligned to 0 with the compute units.
This patch fixes this situation by issuing a warning with a suggestion
to set environment vars and then ignoring the alignment to 0

Per several issues:

Principally:
https://github.com/fireice-uk/xmr-stak/issues/81

Related:
https://github.com/fireice-uk/xmr-stak/issues/490
https://github.com/fireice-uk/xmr-stak/issues/472

increase version to 2.2.0

Make sure we are using a steady clock

fix static compile with non gnu compiler

Stop UAC popup for elevated ctx

remove adding compiler flags `-static-...` if compiler is not a gnu compiler

Add altcoin list

for CentOS, do scl without sudo

Update tuning.md

Link windows to anchor

Update CMakeLists.txt

I ran into a problem following these instructions, and this change got me past it.

The command in question enables access to the C++ compiler (puts it on the PATH or something like that).  If you do it with `sudo`, it leaves you at a root shell prompt.  If you then exit that root prompt, to go back to non-root, then g++ is no longer accessible.  If you want to compile as non-root, then you need to do `scl` as non-root.

The github online editor is injecting tabs for some reason, fixed them.

Adjust spacing and pathing based on review feedback.

Fix typo in socket.cpp

fix windows compile with CUDA 9.1

readd appveyor xmr-stak runtime test

handle cuda error codes

fix cuda9.1 compile

Installed xmr-stak on fresh Ubuntu 16.04 with just AMDGPU-PRO and required these changes and use of OpenCL_ROOT=/opt/amdgpu-pro for successful compilation with OpenCL. Also requires the opencl-headers package from Ubuntu's repository.

applay cmake patch for CUDA 9.0 also for CUDA 9.1

- fix cuda9.1 compile (remove includ eof device_functions.hpp/ removed with cuda9.1)
- remove NVIDIA Volta gpus for MAC OSX

Also fflush in printer::print_str() if configured to do so

readd test removed in #179

handle all error codes from the cuda api calls.

fix windows compile

solve #440 thanks @true-datura

Fix compat_usleep for WIN32

Makes casting more explicit.

When compiled with VS2017, the negative applied to the uint wait
time is ignored. Fixed by casting first.

add binary checksums for version 2.1.0

conservative NVIDIA auto suggestion