Misc failsafe changes

[dktcoding]

This are another unsorted changes, most of them have a reason to be:

• Matrix#equals() wasn't checking for type or null
• Since we store Matrix instances in a Map<String, Matrix> when serializing multiple matrices, we need a proper implementation of Matrix#hashCode() to avoid problems when deserializing in different libai instances (yes, the key of the Map is a String and not a Matrix, but not all maps are deserialized the same way).
• Lot's of resources weren't being closed (the MySQLDataSet was mostly left alone, but no statement or ResultSet is being properly closed, I didn't touch it, because it uses the ResultSet itself inside the iterator and didn't have enough time to properly test it)
• I will really start thinking about a NNIO class
• Add serialVersionUID to all serializable classes (I files serialized with libai compiled with OpenJDK don't work in versions compiled with OracleJDK because they lack serialVersionUID... I tried... sigh)
• About the @Override annotations, since I didn't write libai, it's fairly difficult to follow inherited methods (I'm sure that I'm not the only one).

[dktcoding]

About the @Overrides, I honestly started finding them useful when reading what other people wrote.

Regarding ResultSets, closing them becomes important with most cached drivers (like the latest driver for MariaDB), and when using connections pools where Connection#close() actually releases the lease on the connection but doesn't actually close it.

[kronenthaler]

I will check the resultset issues when i'm checking the DataSet refactors.