I had forgotten about that ordering thing: depending on the platform
cfspack doesn't read files in a directory in the same order.

Some compilers don't like assigning stdout statically.

This allows us to get rid of the zasm.sh wrapper.

Use it in shell instead of using popen()

Also, removed leftover in runbin.c (what did it do there??) that
could result in tests falsely passing.

Things are now much simpler.

HL, instead of being preserved, is set to the character following
the last read character.

Also, make parseHexadecimal "tail" (HL). Soon, all routines in lib/parse
will do that, making the life of lib/expr easier.

This unit is being removed.

Sub-parsers are seldom used by themselves, except for parseDecimal.
I'm tightening the code of this unit for two reasons:

1. Optimization
2. Upcoming API change where HL won't be preserved anymore, but will
   point to char following the last parse char. This will allow us
   to simplify lib/expr.

Finally getting rid of this bad mistake of using IX for this.

Instead of going left and right, finding operators chars and replacing them
with nulls, we parse expressions in a more orderly manner, one chunk at a
time. I think it qualifies as "recursive descent", but I'm not sure.

This allows us to preserve the string we parse and should also make the
implementation of parens much easier.

For some reason, I've mistakenly disabled tests in test_expr without noticing
and I also broke "-123" parsing. Fixed.

Those replacements were trivially equivalent. For the other ones, an
examination of the context is necessary.

This should make tests a bit more convenient to write and debug.

Moreover, begin de de-IX-ization of parseExpr. I have, in a local WIP, a
parseExpr implemented using a recursive descent algo, it passes all tests, but
it unfortunately assembles a faulty zasm. I have to find the expressions that
it doesn't parse properly.

But before I do that, I prefer to commit these significant improvements I've
been making to tests harness in parallel of this development.

This completes instruction support for the Blink tn45 example.

Take argspec pairs directly as an argument instead of taking a
pointer to an argspec pair.

The instruction table is now beefier and has redundant data, but
this simplifies processing. This should simplify upcoming work.

Also, add a "real world" example in AVRA tests, a blink program on
a ATtiny45. Some instructions are commented out because they aren't
implemented yet, but not many.

The output of the program has been verified against AVRA's own
output.