To support multiple objects with differing field values for the same class
type, it is now necessary to actually parse segment 4 and act accordingly.
Thus do that, and add a test to ensure the behaviour is correct.

Due to the way MATLAB stores field names for structures, individual items
in the structure don't have names.  Thus, one must use the field name.  Fix
the grab bag parsing to do this correctly, and add a test to verify correct
behaviour.

Don't mislead in the documentation.  When reading 24 bytes, 24 bytes are read.
Not 16.

Any class that has default attributes only is now fulled loaded and parsed.
Other attributes are currently skipped.  Also commit in a test for this
behaviour that uses a char array as a single parameter.

To avoid future surprises, make sure that the MCOS header bytes for an object
are correct.

For the future, the class name will need to be fetched before I know about
the placeholders.  Thus read in the class names and use that now.

For applications parsing the MAT file later, they may need to know the byte
order of the original file.  Thus store that information in the header.  Note
for writing it is possible to change the endian indicator without changing the
byte order returned.  This should eventually throw an exception.

Verify everything continues to work with multiple elements.

To match the new Group Id, switch the package name.  This also moves the
package name under a controlled domain, avoiding potential name collisions
later.

Fix some comment misspellings and formatting, and remove unused imports.

MCOS is handled completely separately through a weird format.  Thus, start
supporting it by reading in the necessary data.  Currently it isn't fully
parsed, as the first unit test doesn't require extra data.  Also include the
MATLAB class file used to generate the test.

Files produced by Simulink, along with embedded MAT files inside of MAT files
don't have the full header.  Thus enable parsing them as necessary.

Get this going for sanity checks.  Test against both JDK 6 and 7.

Even though most users won't run against this, in case multiple empty keys are
found, store them all under @? where ? is a 0-starting index.  Also include a
modified MATLAB file to test these conditions.  Note that MATLAB just ignores
all three inputs.

Also ignore the IntelliJ IDEA files.

Remove some old tests file that are not used, along with a pre-compiled jar
file that came from somewhere.

Fix up the pom file, preparing it to be used to deploy to Maven Central.
Also fix up some whitespace issues, and lock the versions of all used plugins,
since maven couldn't apparently figure out which versions to use.

Added an equivilant function to getIR() in MLSparse for columns. I thought getJC would do this but looking at the code it did something else. Therefore I added getIC (this is possibly named badly) which literally gives you the column index as an array whose i_th value is the column of the i_th entry in pr. I've also changed the pom.xml so that its version wasn't snapshot. This was so i could create the artifact and serve it off our nexus