Improve User Guide

[rhornung67]

Is your feature request related to a problem? Please describe.

There are many ways in which RAJA User Guide can be improved. This issue captures things we can do to improve it please add more items to the list below:

• Add cookbook section(s) with examples illustrating how to use various RAJA execution policies. Include discussion of performance benefits of one policy over another for basic kernels and show "native" implementations in CUDA, for example, for context.
• Provide more descriptive explanation of execution policies in general.
• Do we want to stress passing resource to kernel execution methods? If so, update discussion and examples to reflect this.
• Add link to the raja-suite-tutorial from raja docs
• More clear and obvious warning about how raja::forall can't be used for nested loops (which is more upfront and easily noticeable - what is there now is easy to miss and i've had a couple new raja users try to use nested raja::forall constructs)
• Add warning or note that for most situations a global threading policy is adequate