Adding the `Fuzzer`class, `NodeRange` -> `span`, and WF contexts.

[matajoh]

This commit adds a new Fuzzer class, which encapsulates the logic from Driver and allows it to be applied to the new Reader, Writer, and Rewriter classes. Another global change is the addition of a stack of wellformedness contexts. By calling wf::new_context and wf::end_context, it is possible to have nested Trieste pass ranges execute. This allows a Trieste pass to execute another Trieste pass as part of its execution (e.g. to parse a string using the JSON parser and the convert the resulting AST to a language-specific format). Finally, this commit changes NodeRange to be a std::span (NB an equivalent class is provided for C++17).

YAML

• Fixed a bug with block indents that contain YAML or JSON
• Fixed a bug where a colon is immediately adjacent to the start of a flow construct
• Fixed a bug with colons immediately adjacent to values inside flow constructs
• Fixed an overeager error message for comments that look like mapping keys
• Fixed a bug with bad aliases in to_json

JSON

• Added unescape and escape functions for handling JSON strings
• Added a symbol table to Object and a Key token to make JSON ASTs easier to use

Misc

• Added some code to mitigate situations in which P(c | d, p) is uniform during fuzzing.
• Added find_first and contains methods on Node which implement Token specific logic.
• Added postparse and output_wf properties for Reader
• Added input_wf, output_wf and passes properties for Rewriter
• Added input_wf and passes properties for Writer

[mjp41]

FYI, I didn't review the json/yaml/infix changes.