This reverts commit 745c6c6d.

Rename InvalidationResultInfo to InvalidationResult and remove Period::getAllParentPeriods and replace its use in ArchiveInvalidator w/ some small SQL changes.

Fix regression, make sure if period is not supplied to ArchiveInvalidator then all periods are invalidated efficiently.

Use Common::prefixTable instead of storing tables_prefix as property. (Don't flout convention for LTS release.)

Adding diagnostic command that allows you to see exactly what's inside an archive table. Useful when trying to find out why one table is using a lot of space or is using more space than others.

Do not perform an inner join on the archive table, for large tables this can make Piwik completely unresponsive and can result in invalidated archives never being purged.

Use mock RawLogDao instance in PrivacyManagerTest instead of test events. Also allow IntegrationTests to override DI config for tests.

Remove ability to specify table in logs:delete command, visit deletion will cascade. Implement LogPurger service using different strategy from LogDataPurger (which can result in error).

Use forEach function instead of iterator to loop over logs and replace in usercountry:attribute command. usercountry:attribute tests pass.

Refs #7181, add logging to ArchivePurger service + archive purging scheduled tasks + Scheduler, modify CronArchive to run scheduled tasks within same process as core:archive so logs are visible in core:archive output.

A chunk implementation that is much simpler and makes more sense. Everything is now in the ArchiveWriter + Selector

* Store subtables in chunks of 100 subtables per blob. Those 100 subtables are stored
  serialized as an array: array($subtableID => subtableBlob). The first 100 subtables are
  stored in "chunk_0", the next 100 subtables are stored in "chunk_1", ...
* Subtable Ids are now consecutive from 1 to X
* We do no longer serialize the whole Row instance when archiving, instead we only
  serialize the Row's array which contains columns, metadata and datatable. This is not
  only more efficient but allows us to refactor the Row instance in the future (although
  we will always have to be BC)
* Faster row implementation: Columns, Metadata and Subtables access is much faster now

Refactor AttributeHistoricalDataWithLocations for clarity, fixing IntegrationTestCase issue and fixing changes to RawLog...