Allow `{` to appear in a concat statement string

[alexlafroscia]

As part of implementing concat_statement parsing, I had to take a short-cut that helped to disambiguate a string as an attribute

<div class="foo"></div>

From a "concat statement"

<div class="bar {{foo}}"></div>

This shortcut was to tell the parser that, when used as an attribute value, a string cannot contain a {. This allows the grammar, when encountering a {, to know that we're in fact parsing a concat_statement instead.

There are a few things we could to do resolve this:

1. Remove string_literal from the possible attribute values

   This would resolve the issue by removing the ambiguity while parsing, by removing one of the possibilities. It should be safe from a syntax-highlighting perspective, because we can highlight the concat_statement as a string anyway. However, the grammar would be "less correct", which may or may not really matter to anyone.

2. Re-write the _mustache_safe_* string literal matches to allow { but not {{

Technically the invalid sequence is {{ not just {. However, I struggled to write a regular expression that would match a string only if it did not contain {{ together in a single sequence, since the "contents of a string" regex is based on characters that are invalid, not ones that are valid. When using the character set exclusion regular expression syntax, you provide a list of individual characters; you can't put two together in there that only preclude the regex from matching when found in-sequence. Some other kind of regular expression would be needed.

3. Move scanning for _mustache_safe_* string literals to C

We would have more control over how the file is parsed in C, where we can use the lexer to go character by character and identify the string as a concat_statement when we find the {{ in sequence.

---

Overall, this would be nice to fix, but is in my opinion low-priority because it will only confuse the parser when an element has an attribute value of a string with a single { in it, in which case it should still be highlighted as a concat_statement anyway so things will look correct. I'm not even sure that this bug is impactful enough to be worth fixing at all!