Splitting a word into MSB/LSB pairs happens often and is worth, I
think, native words. Also, I'm going to need it in the upcoming
commits.

I wanted to make CVM's forth use the Grid subsystem, but doing so
would break the stage binary. Hence, this split.

There are many situations where it can be useful. Worth it.

With KEY and EMIT being switch words, most of the high layer can
be defined before drivers.

In addition to this change, I've compacted core blocks which were
becoming quite sparse.

Adding a delay such as the 20ms one we have in AVR programmer's
initialization routine is tricky without a word like TICKS.

This implementation is highly inaccurate, but more accurate and
reliable than a "ballpark" DO..LOOP...

The few extra bytes they save in the core aren't worth the extra
complexity. This was initially done in a context where I had
troubles keeping the RC2014 binary with SDC inside the 8K limit.

At this point, even with the few extra bytes we add here, we're at
7200 bytes, so I'd say we're fine.

This marks the end of wordrefs in stable ABI. Nice.

When writing the xcomp documentation, I realized that with careful
threading and by accepting a bit of code duplication in the xcomp
toolset, I could de-stabilize a couple of words.

(n) and EXIT are a bit trickier, but I think it can be done. It
would be nice to get rid of stable wordrefs...

This should help spot bugs due to bad initialization.

There is now no more actual code in stable ABI, only references.
This makes refactoring of this code much easier. For example,
changing IY to BC as the IP register.

Only its jump addr remains.

Only its jump at 0x33 remains.

I've also fixed a strange offset oddity in 8086's (n) placement.
It was off by 2, but strangely, it ran properly. Anyway, now it's
fixed.

It's more descriptive this way. Also, I'll soon add a new HERESTART
config.