Update Python bridge to properly handled nested expressions and other fixes

[bettinaheim]

... will update this description shortly. Below are some (by now very incomplete) notes.

This PR

• introduces a proper value stack that makes sure that each node in the AST gets the values associated with it
• makes the value stack/visit automatically validate that are values were processed and that no node pushes more than one value
• changes range and enumerate to create a proper stdvec rather than pushing an array and size as two separate values to the value stack
• changes the attribute visit to not process built-in noise channels; these channels can only be used within calls to apply_noise and are handled there to simplify code and avoid pushing multiple separate values for the same node
• cc.Arrays are now exclusively used only as the underlying data of vectors and for strings; currently, no node visit will produce and array. If that is changed in the future, a bunch other other node visits will need to be updated to handle arrays appropriately.

Bug fixes:

• Fixes an issue where the length of some ranges was not computed correctly, which can lead to segfaults
• Fixes and issue where boolean expressions with multiple operands (a and b and c) were not correctly processed past the first two operands

Other clean up:

• conversions to bool are handled by a shared helper function
• cleaning up loop handling; don't attach 'invariant' attribute for loops that may not be invariant and eliminate code duplication in visit_for and visit_while

TODO:

• enable tests that previously weren't being processed properly
• write a proper set of tests for boolean expressions
• it would be nice to isolate logic around qvec, stdvec and arrays into helpers that the visits call to make it easier to change.

[1tnguyen]

It seems like the GitHub diff UI is choking here as well.