Automatically scale memory based on sample size

[andrew-slater]

Past evidence has indicated that when > 400 subjects in a single analysis/group, qmake chunk job memory requirements could exceed the currently hard coded 3 GB in tableinput.R. Ideally, tableinput.R could automatically scale the memory setting based on the number of subjects in the largest analysis/group (or perhaps it is better for the gtx package to do this as it has more knowledge of the analysis/groups). As a simple fix, allow a config key where the analyst can specify the number of subjects in the largest analysis/group and tableinput.R would recognize and scale the memory setting accordingly. Something like -l mt=NG where N is (X / 100) where X is the number of subjects - unsure if N can be float or if it should be rounded.

[andrew-slater]

The potential challenge with adding this to the gtx package is that the make call is abstracted to use either make directly or qmake (or lsmake etc.) so it would need some logic to detect if qmake is being used and then set mt=NG or if lsmake is being used then set the appropriate option or if make is being used, then don't do anything.