Improve performance by batching inserts

[noschinl]

By batching inserts, performance can be greatly improved.

On my machine, runtime for importing Spain_Portugal.db went from 3m00 to 1m00. I verified that the resulting db contains the same elements by comparing the results of the query

select p.id, p.type, p.name, FR.name, FS.name, TK.name, TV.name
from Points p
         join Points_Root_Sub PRS on p.ROWID = PRS.Points_id
         join Points_Key_Value PKV on p.ROWID = PKV.Points_id
         join FoldersRoot FR on PRS.FoldersRoot_id = FR.id
         join FoldersSub FS on PRS.FoldersSub_id = FS.id
         join TagKeys TK on PKV.TagKeys_id = TK.id
         join TagValues TV on PKV.TagValues_id = TV.id
order by p.id
;

on the db before and after the change.

[noschinl]

This patch does not handle the append mode correctly, you will get duplicate Ids (actually, what is the use case for this mode?). I'll post a patch soon.

[noschinl]

Fixed the bug above; added two changes:

• Always use node_id 0 with Poi_Importer (the ID is supposed to refer to the OSM node id, which is not available. Using 0 seems less confusing than an arbitrary, but wrong id).
• Cleanup of the previous commits.