Compaction Accounts for Continuous Assignments

[calebmkim]

Fix #1767. Adds analysis so that compaction accounts for continuous assignments.

The analysis here is conservative, since it essentially treats all the continuous assignment as one big "block". In other words, any time a group affects the continuous assignments, then it will depend on any previous groups that affect on continuous assignments too. For example say we have

seq {
  A; // Affects continuous assignments
  B; 
  C; // Affects continuous assignments
  D; 
  E; // Affects continuous assignments
} 

Then we detect an A->C->E dependence, in addition to the existing dependencies we already find by analyzing the groups/control.

When I say a group "affects" a continuous assignment, I mean that the group reads or writes to a cell that continuous assignments read or write from (except in the case when the group and continuous assignments are just reading from the same cell; that's the only time that doesn't count).

Also, I checked and this doesn't affect the polybench results from our paper.

[rachitnigam]

LGTM! It occurs to me that we should sit down at some point and review all the analyses and restructure them a bit to be more reusable across passes. The MO with ReadWriteSet has been adding ad-hoc methods every time we have new need.