Support for changing indexed columns on append

[osopardo1]

The current implementation of the Qbeast Format on Spark does not support appending new data with a different indexed columns that the ones specified at the beginning.

Since data is constantly evolving, and business needs may change, it does makes sense that Qbeast index goes in the same direction. Also, it is designed in a way that it can support different index versions (called Revisions), and querying each version independently.

The to-do's of this issue are:

• Remove constraints that checks wether the columnsToIndex are equal to the current indexed columns.
• Check Spark Plan to build the qbeast_hash accordingly. When we remove totally the hash (as proposed in Overhead of qbeast_hash filtering when doing a Sample #68 ), this would no longer be needed.
• Test and check the correctness of the results when doing sampling, filtering and reading the whole data.

[osopardo1]

On hold due to changes on #175 and #68 . Closing both issues would affect directly to this implementation.