migrate from yaml-ast-parser -> yaml package

[cowboyd]

Motivation

The module we were using the parse strings into yaml syntax trees was not maintained, and had an awkward API. This meant that implementing new features which dependend on YAML serialization and deserialization were destined to be difficult. This includes implementing the print function (#51) which is on the hot path for the first iteration of an executable program.

Approach

This uses the npm:yaml module (https://github.com/eemeli/yaml) as the basis for converting strings into yaml AST. To do this, it uses a "read" function which does a 1:1 mapping of yaml ast values to PlatformScript values. the read() operation does not do anything beyond this. So, for example, read("$a-ref") will be a PSString, rather than a PSRef.

There is a recognize() operation that converts the "literal platformscript" value into one that contains references, functions, and templates. The "evaluation pipeline" now looks like this:

pipe(read, recognize, eval)

Mostly this change is limited to the read, however it does have some small implications for the higher level apis because now references are "recognized", not read directly, and so a reference value is not a JavaScript string, but rather a PSString. The same applies for function definitions. We now store the head of the function as a reference to the PSString that gave rise to it. So, for example, in the lambda:

(x)=>: Hello %(x)

The head is the PSString (x)=>

💡When we have macros, we will probably insert that after recognize, but
before eval. e.g.

pipe(read, recognize, macroexpand, eval)