Refactor write page filters for data pages

Refactor monolitic xtrabackup_copy_datafile to split out the write page filter abstraction.

For a given data page determines if it should be written to the backup output. The filters will be: 1) pass-through (i.e. no filter) for full backups; 2) (slow) incremental backup page filter for incremental backups that must fully scan the source data files. 3) compact backups filter.

The motivation for these abstractions besides cleaner code is to add the following filters in dependant BPs: the compact backups page write filter to skip over secondary index pages and to have a base code for further read filter abstraction.