Fix for #149

[wseaton]

This patch solves the issue outlined in #149 on my particular database, could probably be refactored to reduce repetition but couldn't figure out how in a way I liked, any pointers? 🙂

[wangxiaoying]

Thanks @wseaton ! The code looks good. Can you also add this modification to the BinaryProtocol's prepare stage: https://github.com/sfu-db/connector-x/pull/153/files#diff-9a26d294fe1517a58485b5b2709196a4613e51fd9f99acd82ca16028a771b53eL218 to make the count consistent? (It seems like we didn't add the check limit on CSVProtocol and CursorProtocol previously, can you also help to add this check? Thanks for helping to find this inconsistency!)

The repetition looks ok to me since it's not much. But if we want to reduce it, I think one way might be extracting a count function, which's input is reference to a connection and a query, and output is the count result in usize. All three protocols can invoke it in this function prepare stage and set the self.nrows using the returned result.

[wseaton]

@wangxiaoying was above to remove the duplication as per your suggestion, still works great against my database. Thanks for the help!

[wangxiaoying]

@wseaton Looks great! Thanks for the PR. I have merged this.