This class exposes two functions. `evalNode` take an AST node and evaluates it down to simplest terms. `eval` takes an array of AST nodes and calls evalNode on each of them. I don’t think we’re quite ready to be producing HTML yet so let’s not get ahead of ourselves.

Currently I'm relying on JavaScript's semantics
around truthiness. I think I'm going to retain
that ease of use (i.e. `(if list #t #f)` should
return true if the list has more than 0 elements,
etc) but I want to specify that behavior myself.

Just numbers and strings for now

I noticed that if you passed in just a number, JS
would helpfully parse that and then
`source.length` was undefined, which broke the
lexer.

I keep test .oslo files around and I hate not
having a clean repo.

I added upper and lowercase because that seemed
preferable to always casting every html function
name to lowercase.

It's a little hacky and not very clean but it's
working for now.

The way that we were doing it before was going to
make it very difficult to implement any sort of
error handling or extension to the language.

Because I was assuming that every expression
started with a literal, calls to `each` were
getting wrapped by a functionCall node with no
functionName, which was subsequently ignored by
the compiler. This removes all of that uselessness

What was happening was that spaces were allowed in
all literals to account for quoted strings, but
this meant that two identifiers in a row would get
parsed as a single literal. This commit sets a
flag to allow whitespace in literals when a quote
is encountered and then unsets that flag at the
next quote.