Data count section is insufficient for shared data segments

The point of the data count section is to pre-declare the number of data segments so that the code section can be parsed to refer to pre-allocated segments. The code should be able to be generated and then the segments should be filled in once the data section is parsed and everything should just work. But there's not currently any way to know in advance if a data segment will be shared, so if the code needs to be generated differently for shared segments, we're plumb out of luck.

It seems we need to be able to emit the data section before the code section or introduce a beefier version of the data count section that can declare data segment types.

Start by reading the data count section, code section, and data section discussion in the proposal. Trace how shared data segments are currently represented and how generated code refers to pre-allocated segments. Done means agreeing on and specifying a way to determine shared segment handling before code generation, either by ordering the sections or extending the data count section.