io/metadata: increase `csv.field_size_limit`

[joverlee521]

Prompted by nextstrain/seasonal-flu#324 and discussion in Slack.

We could increase the csv.field_size_limit in io/metadata.py and/or add an envvar to set the limit (e.g. AUGUR_FIELD_SIZE_LIMIT.

[victorlin]

Given that it was indicative of a data issue in nextstrain/seasonal-flu#324, the fact that it errored instead of silently continuing could be considered a feature. How about this flow?

1. Run with default csv.field_size_limit

2. User sees a helpful error message like:

   The value in row x, column y in file z is too large for Python's default CSV parser.
   Please check your data. If the data is correct, re-run with <env var>.

[joverlee521]

Yes, it would be nice to have the error message for digging into the data issues!