this commit updates the parser and grammar with new syntax for the data
section. it also adds cases for raw bytes (which are generated by
`--optimize codesize`) in the data section, which are required to
round-trip venom produced in `--optimize codesize` mode.

new format:
```
data readonly {
    dbsection foo {
        db @label1
        db x"abcd"
    }
}
```

the new "readonly" modifier doesn't have disambiguating power yet, but
it will make it easier to add "writable" data sections to the grammar in
the future if we decide to add that.

misc/refactor:
- remove xfail from previously failing test cases.
- refactor internal representation of data section to not use
  `IRInstruction`
- add `_as_asm_symbol` for easier lowering of labels
- tweak `IRFunction.__repr__()` to add the function name at close
  (helpful for debugging).
- tweak `IRBasicBlock.__repr__()` to add basic block and function name
  at close to help with readability