Summary:
**Problem**: There are errors occurring in named tuple for a special case where the named tuple fields contains commas (or multiple spaces).

Ex. G = namedtuple("G", "a, b, c")

**Root cause**: In the preprocessing.ml file: the preprocessor currently splits the field name by single whitespace but this does not account for commas nor repeated whitespace.

**Fix**: I implemented a change to split the field names by either comma or space. This is followed by filtering out all resulting stray empty strings that result from the splitting process. This approach has been tested by multiple different cases and holds up to combinations of commas and whitespace.

**Alternate approaches considered**:

1. Implementing a regex expression to alter the input string.

Reason against approach: it could involve some pretty complicated regex logic: I would have to either check for a single comma followed by 0-infinite whitespace OR no comma followed by 0-infinite whitespace. Other combinations such as multiple commas would be invalid.

2. Implementing a parser that iterates through the string, checks for invalid field name strings and throws errors.

Reason against approach: The python language does NOT throw error at invalid field name string. I didn't want to have different logic implemented by the language and the pyre typechecker.

Thus, I decided to go with the approach of splitting by either comma/space and filtering out empty strings.

Reviewed By: shannonzhu

Differential Revision: D36138262

fbshipit-source-id: b3cac128ae9d6e1ddfa03e09597ce64713758ff7