Row normalization

[Unlink]

Closes #36 and #100

[dg]

IMHO $cols[$meta['name']][] = $tableName; is always good, without if

[dg]

For what cases it is necessary to have its own normalizer? And normalizer and own row factory?

[Unlink]

Sry for late response. So, at first this PR allow create your own row normalizer for whole session because you should specify it in config. Sometimes i need work with geomatics data types eg. point and it will be nice if i could wrap it in my own class globaly. According to #36 i added possibility to specify row normalizer for specific ResultSet, so you can for example turn off normalization if you need.

RowFactory is used internally in Selection. It's because i want to use whole ResultSet with duplicate columns check feature... so i don't create at first Row instances and then ActiveRow from them but i used row factory so ResultSet::fetch return correct instances. I was also thinking about change ActiveRow to use Row instead of plain array, but i don't know.

[dg]

@Unlink Thanks for all the PR. I'm sorry that I did not have time for Nette Database. I merged commit with improved error message on duplications. Now I'd like to release NDB 2.4, because it is the last unreleased package, and than I'll merge others PR like #102/#105, row normalizer, and we can discuss about refactoring Structure. Next NDB 2.5 can be released in few months. Ok?

[Unlink]

ok, thanks :)