This is a huge release, with the parser at 67MiB in size. But we gained
a lot in return.

Starting with parse statistics:

Total parses: 3987;
successful parses: 3939;
failed parses: 48;
success percentage: 98.80%

We are now very close to parsing 99% of the Nim compiler, and that's
including the tests where parsing is supposed to fail.

Let's dive into the big changes:

Breaking changes

String and comment content has been spun off into their own nodes
(#54). In addition to this, string_literal has been specialized into
interpreted, raw and long variants. This should allow for a much
easier time writing capture queries.

Parsing improvements

Statements and blocks now terminate as close to their origin as
possible. This means:
```
foo()| <- terminates here

bar()
^ instead of here
```
Layouts are now ended whenever indentation falls off, which improves
tree-sitter ability to figure out what error happened.
The scanner now properly terminates statements in error recovery,
which allows for less syntax tree unraveling when such events happens.
"Terrible" if/try statements are now handled properly. The
statements in question usually look like this:
```
if x:
  y else: z

try:
  proc foo() =
    something except: bar()
```
These can be found within the Nim compiler source code.

Relaxed rules for generic parameter list and tuple field list. The
following are now valid:
```
proc foo[]()

tuple[]
```
This is done to align with how the Nim compiler handle these cases.
Multi statements term rewriting patterns are now supported.
Full support for Unicode identifiers and operators. This support was
previously removed in 0.0.2, but recent improvements have allowed them
to return.