Targeted HDU/Data creation from Java data objects

[attipaci]

In 1.18 we have deprecated the [...]HDU.encapsulate(Object) methods, because they were not safe beyond internal use by the library itself. However, some similar methods could be useful for the targeted creation of specific types of HDUs from data, rather than using the automatic HDU wrapping by FitsFactory.makeHDU(). For example, a float[][] array may constitute data for a 2D image, or else it may be a column major table data (for either a BinaryTable or an AsciiTable). Fits.makeHDU() will always wrap that data into an ImageHDU, but sometimes that's not the intended way to represent such an array.

Here are some ideas for new methods:

1. static <Data-subclass>.from(Object) methods to wrap data into particular Data subclasses, when it can be done in an unambiguous way. (BinaryTable has disambiguated fromRowMajor(Object[][]) and fromColumnMajor(Object[]) methods instead, and AsciiTable should as well...)
2. Data.toHDU() method to then encapsulate a data object into a corresponding HDU, whose header is populated with the mandatory minimal description of the data automatically.

As such, you could have:

   float[][] values = ...

   ImageHDU im = ImageData.from(values).toHDU();
   BinaryTableHDU btab = BinaryTable.fromColumnMajor(values).toHDU();

[attipaci]

It turns out that ImageData had no checks for what object it was created with. The constructor worked with no complaints with any Java object. This is clearly not a good idea. We should introduce checking that the object is a proper primitive numerical array (possibly multi-dimensional)...