Build a parser in C

[rmosolgo]

Like #4366 , this ports the existing parser (racc) into a C extension (using bison).

It looks like it's about 10x faster: After improvements to the Ruby scanner, the C parser is 7-9x faster:

  ~/code/graphql-ruby $ be rake bench:parse
  # ...
  parse - introspection
-                           1.243k (± 8.4%) i/s -      6.174k in   5.005459s
+                          11.068k (± 2.8%) i/s -     55.650k in   5.031916s
-    parse - fragments      1.870k (± 7.0%) i/s -      9.353k in   5.029178s
+    parse - fragments     14.839k (± 6.1%) i/s -     75.168k in   5.086781s
-    parse - big query    222.342  (± 3.6%) i/s -      1.122k in   5.053040s
+    parse - big query      1.670k (± 5.3%) i/s -      8.496k in   5.103380s

As well as less memory:

- Total allocated: 498928 bytes (8187 objects)
+ Total allocated: 318544 bytes (3570 objects)

• Changing the token type from TYPE to TYPE_LITERAL breaks the Ruby parser, fix this upstream
• Add SDL parsing to the C parser
• Extract parsing tests to a module and share them between the parsers doing this in a build instead.
• Try https://github.com/Shopify/ruby_memcheck
  • Moved to Add RubyMemcheck to CParser #4410