module enhancements part 6: sampling pseudo-absences

[CeresBarros]

@CeresBarros has noticed a poor statistical performance when the ratio of fire presences/absences is small (e.g. 205 fires across 29 years in a total of 19295 pixels, ratio = 205/19295 = 0.01062451).
A solution to this is to sample pseudo-absences instead of using the full background data. A few questions arise from this:

• should we warn the user about small ratios? How small?
• should pseudo-absence sampling always be done/advised?
• when doing pseudo-absence sampling, the predicted probabibilities will need to be adjusted.
  - How?
  - How to account for no-fire years (that only have absences)?
  - This adjustment needs to be "communicated" to fireSense_ignitionPredict.

Further detail here