Improve string names for locally-bound Heapster variables

[eddywestbrook]

Currently, whenever a new local name is bound in Heapster, it is called either "zNNN" or "xNNN", depending on how it gets introduced, where "NNN" is a unique number. This PR changes that so that the variable name starts with a prefix which matches the type; e.g., LLVM pointers become ptrNNN, bitvectors become bvNNN, etc.

[glguy]

The change looks good.

Things worth considering, but not blockers:

• When setVarType is given a (Just str), use that as a prefix/suffix along with the typeBaseName so that top, local, and ghosts benefit from their type-based name-hint as well.
• It's probably time to start putting {- ^ ... -} -> haddock comments on the type signatures with multiple Maybe String arguments.

[eddywestbrook]

Awesome, thanks Eric! I'll implement those changes.

[m-yac]

Looks good to me! I agree with Eric's comment about adding comments.

I have one very unimportant observation: The error case of ppInfoAddExprName seems slightly strange to me. At the moment it will never be hit because it's always passed the result of typeBaseName, but if in the future we include user-defined strings in typeBaseName and one of those strings happens to have a number at the end, we probably don't want to error, I suspect we really just want to use some default name.

[eddywestbrook]

@m-yac: I see what you are saying, but I don't actually expect those to ever be user-defined strings. The only place that could possibly happen would be from the C variable names, which always end with a close bracket, i.e., always look like C[xxx].

[eddywestbrook]

Merging manually, because the only changes since the CI succeeded last time are additional comments and a slight tweak to what strings are printed.