Steer suggested optimizations based on profile results

[emeryberger]

Is your feature request related to a problem? Please describe.
Right now, Scalene's proposed optimizations focus exclusively on CPU and/or GPU time. We can do better with the information Scalene already has collected.

Describe the solution you'd like
Offer suggestions tailored to the profiler results. For example, if the line or region contains considerable Python memory allocation (for some definition of considerable), propose an optimization aimed at reducing this or using native data structures instead. Similar suggestions would make sense to target reductions in copy volume or to reduce GPU memory. We might also extend Scalene to offer solutions tailored to multicore if the number of cores being utilized is low (or if it is requested in a checkbox under Advanced Settings, as one can now do with GPU optimizations).

[emeryberger]

The repository version now includes radio boxes under Advanced settings that let users choose between performance and memory optimizations, so this can be manually selected.