Add support for per prepared statement query options

[jeremyevans]

I just added support for using mysql2 0.4+ native prepared statements to Sequel (jeremyevans/sequel@5066f16). Based on my experiences with it, I think improvements to the API would be beneficial.

In the general API to send queries to the server (Client#query), you can specify options on a per-query basis. However, the prepared statements API doesn't appear to support that. Client#prepare only takes the sql, not an options hash, and Statement#execute only takes arguments, not an options hash. Statement gets the query options to use by copying the Client's query options inside of Statement#execute. Result stores the query options in an instance variable, but does not offer a method to access them.

To work around this issue, you need to modify the Client's query options before calling Statement#execute, then reset the Client's query options after. This is what Sequel does in order to make sure that the behavior for prepared statements matches the behavior for non-prepared statements in regards to which options are set.

The other possible way to work around this is to modify the query options in the Result, but that has issues:

1. It's ugly (result.instance_variable_get(:@query_options).merge!(options))
2. It causes warnings in Result#initialize if the Client doesn't have :cache_rows=>true (https://github.com/brianmario/mysql2/blob/master/ext/mysql2/result.c#L890)
3. It can't handle streaming, since that happens before the Result is created (https://github.com/brianmario/mysql2/blob/master/ext/mysql2/statement.c#L366-L370)

I think the best way to fix this would be to copy the Client's query options to the Statement in Statement#initialize, and add Statement#query_options. Alternatively, you could add a Statement#execute_with_options method that accepted an options hash and merged it into the Client's query options.

[sodabrew]

Take a look at #640 and let me know if that looks reasonable to you.

[jeremyevans]

@sodabrew That is exactly what I had in mind. Sorry for not searching first, I should have just commented on that issue. I'm going to close this now as it is a duplicate, but we can continue this discussion if you want. Is there a reason you have been hesitant to merge #640?

[sodabrew]

I wasn't sure if it was actually needed, or if it even made sense to store different query options into each statement. It's also not obvious to me whether a change to query options on the Client itself should override the options on a Statement that was created previously.

[jeremyevans]

I tend to favor the copying approach, so that after the Statement is created, future changes to the Client's options are not picked up, but you could always just merge the Statement's options into the Client's options on every call to Statement#execute if you want it to pick up changes in the Client's options.

[kamipo]

I'd like to support for per execute query options for Active Record because AR caches prepared statements in statement pool.
Ref rails/rails#25507.

[sodabrew]

So I finally figured this out! Ruby 1.9.3 added : to rb_scan_args. Easily implemented in #912.