The type used to store the index for a case-type now tracks the
type of the index expression rather than always using an "int".
The case fields also now have some checking done at code-gen-time to
ensure the constants used for cases does not exceed the numeric limit
of the type used in the case's index expression. Then, assuming, it
looks safe, the C++ case labels are generated with casts to the type
of the Binpac case's index expression to ensure compilers accept it
(since all Binpac numbers use "int" for storage/printing internally).
This allows for tunability of the following behaviors:
* Minimum flowbuffer capacity to use when parsing a new unit
* Threshold at which flowbuffer capacity is contracted back to the
minimum after parsing a complete unit and before parsing the next
* Maximum flowbuffer capacity to allow when parsing a given unit
Failed flowbuffer allocations due to reaching maximum capacity or any
other reason now throw ExceptionFlowBufferAlloc.
