Issue 304 streamline tune

[evaham1]

In this PR I have edited the exisiting tune functions and added new tune functions so that tune() can now be used to both optimise number of variables to keep and also number of components to keep. tune() functions now broadly have the following functionality:

1. Optimise number of components to keep

• Tells you what number of components are informative for a given model
• Available for all tune() functions apart from tune.spca() and tune.rcc()
• Is calculated by internally building the revlevant model and then calling perf() on this model (only exception is tune.pca() which just returns variance explained per component)

2. Optimise number of variables to keep

• Tells you what number of variables are informative per component for a given model
• Only available for sparse methods, i.e. tune.spca(), tune.rcc(), tune.spls(), tune.splsda(), tune.block.splsda() and tune.block.plsda()
• Is calculated using the code from the original tune() functions

List of current tune() functions and which functionalities they have:

• tune.pca() - (1)
• tune.spca() - (2)
• tune.rcc() - (2)
• tune.plsda() - (1)
• tune.splsda() - (1) and (2)
• tune.pls() - (1)
• tune.spls() - (1) and (2)
• tune.block.plsda() - (1)
• tune.block.splsda() - (1) and (2)
• tune.mint.plsda() - (1)
• tune.mint.splsda() - (1) and (2)

For the functions that have dual functionality, optimising components (1) can be done by setting test.keepX = NULL (and also test.keepY = NULL for tune.spls(). If test.keepX is not NULL, functionality (2) is used.