Support finishing component types without compilation artifacts

[macovedj]

As referenced in this PR, the addition of component type reflection support breaks jco's current dependence on wasmtime-environ, and since jco doesn't leverage an Engine, I'm not immediately aware of a good way to adapt jco. So this PR just does the simple thing and reintroduces the old finish function with the new name finish_sans_reflection, right next to the one that was updated for reflection support, though if there's a way I'm missing to adapt jco, then this could be reconsidered.

[guybedford]

Can confirm this gets the tests in the JCO upgrade PR to pass on bytecodealliance/jco#368.

[alexcrichton]

Thanks! Would it be possible to call finish as-is though? I think passing an empty PrimaryMap and empty iterators should preserve the same behavior for jco?

[macovedj]

Thanks! Would it be possible to call finish as-is though? I think passing an empty PrimaryMap and empty iterators should preserve the same behavior for jco?

Yeah it seems you're right. Just pushed to the jco branch and have it pointing at wasmtime main with empty PrimaryMap and iterators.