Require LACE_STACK=1 to use non-standard 'stack' instruction set extension

[dxrcy]

PUSH, POP, CALL, and RET instructions will now cause both an assembler error and a runtime panic if used without enabling the stack extension.
The stack extension can be enabled by setting the environment variable LACE_STACK=1.

Example:

main
    call hw_sub
    halt

hw_sub
    lea r0 hw
    puts
    rets

hw .stringz "Hello from the stack\n"

If ran with lace run stack.asm, this program will fail to assemble, with the following error:

Error: parse::stack_extension_not_enabled

  × Non-standard 'push' instruction used without 'stack' extension enabled
   ╭─[2:5]
 1 │ main
 2 │     call hw_sub
   ·     ──┬─
   ·       ╰── non-standard instruction
 3 │     lea r0 sr
   ╰────
  help: this instruction requires the non-standard 'stack' extension
        run with `LACE_STACK=1` to enable
        note: this identifier cannot be used as a label

If ran with LACE_STACK=1 lace run stack.asm, the program will assemble and run correctly.

If a program assembled (with the stack extension enabled), is then separately ran (without the extension enabled), the runtime will print the following error, and immediately halt:

You called a reserved instruction.
Note: Run with `LACE_STACK=1` to enable stack features.
Halting...

[rozukke]

I think instead of doing this in the parser, everything can be done in the check_instruction fn in the lexer, where the effects would be contained in that function entirely (disregarding the env infra, and might have to change return type to Result for a few fns)

[dxrcy]

I think instead of doing this in the parser, everything can be done in the check_instruction fn in the lexer, where the effects would be contained in that function entirely (disregarding the env infra, and might have to change return type to Result for a few fns)

I was originally going to do this but I assumed making the ident and check_instruction methods fallible was not preferable. Changed now.

[rozukke]

LGTM 👍