Add code generation for benchmarks

[mtdowling]

It would be nice to be able to code-gen the benchmarks from a list of expressions. The generated code would then create a normalized benchmark name for each of the following (where TESTNUMBER if the position in the list of expressions):

• lexing (e.g., tokenize_TESTNUMBER_foo_dot_bar_dot_baz)
• parsing (e.g., parse_TESTNUMBER_foo_dot_bar_dot_baz)
• interpret (e.g., interpret_TESTNUMBER_foo_dot_bar_dot_baz

benches could then be generated from either a JSON file or a macro based DSL:

let benches = vec![
    add_bench!("foo", "{\"foo\":true}"),
    add_bench!("foo.bar.baz", "{\"foo\":{\"bar\":{\"baz\":\"bam\"}}}"),
    add_bench!("foo.bar.invalid", "{\"foo\":{\"bar\":{\"baz\":\"bam\"}}}"),
    // ...
];