Double-cells, `BLK@` and faster

• Adds DS+, DS- and DS+! for adjusting double-cell values by single-cell ones.
• Adds BLK@ for reading a disk block into a buffer.
• Optimizes the arithmetic operations.

Faster, smaller, stronger

Several big refactorings here. Same vocabulary and so on, but smaller and faster.

• Now uses direct threading with a smaller, faster NEXT.
• Several words were ported to assembly for speed.
• Ships two separate ROMs: forth_boot.rom immediately runs a disk, while forth.rom is ready for interactive use by a keyboard.

Streaming actually works now

Streaming files only worked for files that fit into one block, in v3. It's fixed now.

Also, added -EX for subtract with carry.

First public release

This is the first released version of TC-Forth.

It supports block and streaming files loaded from disks. It is intended to target the Techcompliant flavour of the DCPU.

It should work on any emulator that supports the LEM1802, standard keyboard, and M35fd floppy disk.

It uses the log and brk instructions added to the TC version of the DCPU-16, but only after an error or when requested by user code, so it should run passably on other emulators.

The forth-boot.rom binary expects a streaming-style disk to be loaded at startup.