improve fgout and fgmax capabilities

[rjleveque]

Some improvements to fgout grids were introduced in #617, but soon I hope to add more, in particular to allow smaller files to be written in situations when lots of frames are output for animations or particle tracking...

1. Allow specifying which components of q to output for each fgout frame.
2. Allow specifying one or more components of aux to be output, and whether to output these every frame, as might be needed for aux variables used to track time-dependent quantities (and/or when the underlying AMR grid resolution may vary with time), or only once.
3. Currently only point_style==2 is supported, a uniform 2D grid. Introduce others as in fgmax, in particular point_style==4, where a mask is specified so that not all points on the uniform grid are output. This will be very useful for reducing the file size if many points are at high elevation, where water will never reach.

For fgmax:

1. Allow specifying which components of q to output more generally.
2. Add additional arrival time capabilities, to keep track of the first time the water depth is above several different thresholds, or falls below various thresholds for drawdown (as requested by WA State).
3. Allow specifying binary output (also binary32) as now supported for fgout grids. Important for PTHA or machine learning applications where large numbers of runs are performed.

Also need to update the documentation,

• https://www.clawpack.org/fgout.html
• https://www.clawpack.org/fgmax.html