{apachesolr_search_node} table should be rebuilt when index is deleted

Currently we use {apachesolr_search_node} to track not only which nodes need to be updated, but also whether there were past attempts to delete a document (e.g. a deleted or unpublished node) which failed. This latter information become irrelevant if we delete the Solr index, so we shoudl rebuild this table fully to avoid making lots of extra irrelevant delete requests.

This is also important for development sites which may be doing bulk deletes of nodes - after such an operation deleting the index is likely to be more efficient than sending all those individual delete requests.