Catch mdb_cursor_renew(write txn's cursor).  Add flag C_UNTRACK, so
mdb_cursor_close need not peek inside a possibly-freed readonly txn.

Close remaining cursors when committing a write txn. The doc says this
happens, and it avoids cursor tracking when updating mainDB + freeDB.

Rename mdb_cursor_merge() -> mdb_cursors_close() for code reuse,
and add a merge option.  Simplify its loop a bit.

Factor out cleanup of DBIs.

On mdb_env_copy() error: Abort the txn.
On mdb_txn_renew0() error: Release new MDB_NOTLS reader slot.

Redo 8a562f56 so !mt_dbxs indicates
"txn was reset", so mt_numdbs gets one magic value instead of two.

Reset me_dbflags[dbi] when closing DBI, to get rid of MDB_VALID flag.
mdb_env_close(): Re-fix DB-name memleak. DBIs > me_numdbs may exist.

The circular dependency issues appear to have been resolved.
Still, need to watch closely, maybe revert this change if
problems arise.

Due to underfilled branch page. We're in the process of merging/moving
nodes to it because we already know it's underfilled. Took this approach
rather than just removing the assert in mdb_page_search_root, because
that assert may yet catch other situations we don't know about.

(Although, it has been there since the original commit of mdb.c and
has never triggered any other times...)

Show MDB_PERSISTENT/MDB_VALID/DB_VALID relationship.
mdb_txn_renew0(): Remove obsolete "cannot fail" comment.

Move key init into mdb_env_setup_locks().
Don't create unused TLS key when read-only filesystem.
Drop internal flag MDB_ROFS, we can instead test either
!me_txns, !mt_u.reader or me_lfd==INVALID_HANDLE_VALUE.

Malloc before I/O. Avoids possible malloc error after I/O.
Don't allocate dirty & free lists when MDB_RDONLY.
Factor out code.

Don't do anything with (fake root) subpages

Fix 227329c8, don't persist
the MDB_VALID bit in db.md_flags.

Need to check NUMKEYS as well as fill threshold, when deciding
whether to change anything. Don't let the page drop below the
minimum number of keys.

This reverts commit 372b3ddb.

Original code works as designed.

mdb_dbi_close() must allow any dbi < maxdbs, since opens in an
uncommitted txn don't increment env->me_numdbs.

This reverts commit 8eef7a42.
Previous commit was correct, duplicate keys should not appear here.

Check for overflow pages, reset cursor position properly.

In APPEND don't immediately reject matching key, since this
is valid for APPENDDUP.

Append mode should *reject* keys that are too small.
Also allow APPENDDUP in mdb_put().

When fixing other cursors, must also fix their depth.

Make sure C_INITIALIZED gets set on successful call

Variables mdb_debug, mdb_debug_start were used undefined.