Because icore.fs defines it in a "not-for-runtime" manner, dummy.fs
executing it had ill effects. We use a workaround instead.

Lifting ourselves by the bootstraps!

It's not needed anymore

Ain't that not self-bootstrapping enough to your taste? Whoa, I'm
getting dizzy...

Now we're seriously getting into real boostrapping territory...

This limitation was becoming annoying...

emul/forth/stage{0,1}.asm now only configure constants and append
binaries.

They do save space (nearly 100b), but they complicate our Forth-ification
process. We'll re-add them later.

This allows us to save a whole 500 bytes on the final binary size!

This change comes after I took a look at the hex dump and saw that one letter
constants in z80a.fs took a lot of space.

The removal of the "divide" routine highlighted the fact that
nativeWord wasn't properly stabilized.

This requires us to significantly adjust our build process, which
now has 3 stages.

Otherwise, when a defining word would be called outside a definition
itself, it would get the name of the last parsed word, that is,
itself!

For example, dummy.fs, instead of creating a "_______" entry, created
a "(entry)" entry...

The goal being to add a new native code dict staging phase.

This should help with upcoming challenges in bootstrapping z80 code
compiled through Forth.