Handling escaped variables in region analysis

[vesalvojdani]

This is meant to address two issues we became aware of about escaped variables and the region analysis.

• Simply handling pthread arguments and treating escaped variables as globals, so we handle regtest 09-34 without resorting to the empty region map override.
• Dealing with escaped owners when a region member escaped. Would be nice if one could trigger an escape event for this, although it would be extremely imprecise because of the non-directional heap abstraction. The alternative would be to re-implement a region-specific escape analysis.
• (Optional) Understand what region owners actually mean: probably one should use AddrOf(i) as owner in 9-34 when its not a pointer.

[jerhard]

Please add a test case for the cases that are fixed with this PR. Then we can consider merging this PR.
The following issues may be related to the behavior addressed here: #107 #1070 #1298 #1300.
Please check whether any of these is affected.