Now that we use xcomp unit, boot entry creation can be a bit
cleaner.

I forgot to do it earlier. NOP having an opcode of 0, it didn't have
any effect.

Should be significantly faster now. Less push/popping and usage of
DE instead of IY.

Saves quite a bit of binary space. More than 300 bytes in forth1.bin

And fix broken negative display in ".".

It is much less confusing when "IFZ," means "If Z is set, continue
into IF, else, jump to THEN,".

Remove flagsToBC boot routine.

Makes boot binary a bit bigger, but that "_pdacc" word will be
reused in high level apps.

Only a few bytes saved in forth1.bin, but the DO .. LOOP construct
isn't used much yet. It's still significant savings per LOOP call.

Allow us to rely a little less on labels. We now only need 4.

It's not used

It makes the tool much more versatile. I'll have adev being included
in all recipes, so it can be assumed.

The former was only used in the peculiar context of "/emul". The
regular case is actually HERE pointing to RAMEND on boot.

This allows optional subsystems to not needlessly reserve too much
system memory and it also reduces offsets bookkeeping.

so that it isn't picked up automatically at boot. Only when (c<$) is
called do we override (c<). This fixes tricky boot problems.

Let's revisit the relinking approach, with LITA this time.
This reverts commit 6652125d47.

It seems that xcomp can't manage to reliably compile core+ units.
Let's limit the role of xcomp to z80c+icore.

I'm getting a little desperate here. I'll revisit the link.fs approach
with a new idea I had...

Replace its usages with direct RAM+ offsets. The (sysv) mechanism was
incompatible with cross-compilation of a full interpreter.