custom tokenizers peliasNameTokenizer & peliasStreetTokenizer

[missinglink]

this PR introduces two custom 'tokenizers'.

the motivation behind this is to deal with cases where the data being indexed contains a non-whitespace delimiter:

# before

"a,b" -> ["ab"]
"a/b" -> ["ab"]
"a\b" -> ["ab"]

# after

"a,b" -> ["a","b"]
"a/b" -> ["a","b"]
"a\b" -> ["a","b"]

this was particularly noticeable for cross streets such as Bedell Street/133rd Avenue which were being incorrectly tokenized as:

{
  "tokens": [
    {
      "token": "bedell",
      "start_offset": 0,
      "end_offset": 6,
      "type": "word",
      "position": 1
    },
    {
      "token": "street133rd",
      "start_offset": 7,
      "end_offset": 19,
      "type": "word",
      "position": 2
    },
    {
      "token": "ave",
      "start_offset": 20,
      "end_offset": 26,
      "type": "SYNONYM",
      "position": 3
    }
  ]
}

closes #91

[orangejulius]

👍 on the code. let's run the acceptance tests in autocomplete mode on this once dev is done

[orangejulius]

Can you confirm that this code is in the dev build that finished today?