- use iverilog's -gstrict-expr-width throughout test suite
- add warnings for excess bits or padding zeroes in number literals
- add new --oversized-numbers parameter to retain support for unsized
  numbers wider than 32 bits
- localized use of oversized numbers to new truncate test suite which
  verifies the behavior of both modes, and compares against the known
  behavior of iverilog

- detect and forbid non-positive size casts
- detect and forbid negative struct pattern literal indices
- avoid invalid constant folding of negative based literals

- standardize number casting
- fix extension of casts of unsized literals to 32 bits

In places where an optional space was previously allowed, allow any
number of arbitrary whitespace characters, in line with the spec and
other tools.

- delegate cast type and sign resolution to TypeOf conversion
- internal support for injecting data declarations into statements
- fix size and sign of unbased unsized literals used in casts
- avoid generating many unnecessary explicit casts
- support casts which depend on localparams within procedures
- expression traversal correctly visits types within type casts
- fix typeof on expressions of net types
- handle additional edge cases for unsized integer array patterns
- preserve signedness of implicitly flattened unpacked integer arrays

- number literals are parsed rather than stored as strings
- fix array query functions used on non-trivial number literals
- more efficient expression simplification recursion
- expanded constant folding scenarios