Implemented sparse QR decomposition

[ZimmerA]

This is the implementation of a QR decomposition using the givens rotation method.

A sum up of the changes:

• Added the function QRSparse which calculates the factorization for a given SpraseMatrix
• Altered the leastSquares function to handle the correct multiplication of the vector with the transpose of Q in case Q is the result of a Sparse factorization. This needs to be done because the Q resulting from QRSParse is a product of givens rotations other than the "regular" dense Q.
• Added DotNetExtions.fs which, for now, extends the .net generic list with some generally useful functions that I needed in order to keep the QRSparse code clean.
• The "QR" function now pattern matches the union type of the given matrix and calls the according QR function (QRDense or QRSparse)

Feel free to give me any kind of feedback or suggestions.

[kMutagene]

Cool stuff! I have a request to you though : could put this pull request on the developer branch? We mainly push changes to these branches first, and then merge them into the master on a semi-regular basis. Sorry if that was not communicated well.

[ZimmerA]

Thanks! Of course, my mistake.