Minimizes the number of memcpy's if additional dups are stored.

Append mode now does no key comparisons, input must be in sorted order.
page_split was not updating cursor parents correctly.

Just lookup the DB roots as needed. When many DBs are in use,
most of the copies won't be referenced in a given txn, and
there's a bad race condition in the copy routine.

Must begin with '/'

This reverts commit 4a0b3049.

Relevant info should be copied during txn_begin only.

Just a trial. This may not make sense if we decide to split the
meta pages into their own separate file, to allow meta traffic to
reside on a separate spindle.

Clarify prev commit, fix Doxygen comments broken by earlier changes

Don't re-use free pages so soon; that leaves us vulnerable to
DB corruption if data syncs successfully but meta doesn't.

Prefix ID, NOID, ID2, IDL, ID2L with 'MDB_'.
Also drop bdb from file description.

Put one call in #if MDB_DEBUG. Move DPRINTF(Using meta page #) there.

Define MDB_FDATASYNC() instead of redefining fdatasync().

If the database is opened with MDB_RDONLY or MDB_NOSYNC,
instead use the same file descriptor for me_mfd and me_fd.

Also factor out Windows/Unix error handling after open.

Get txn->mt_txnid from the meta page instead.  Otherwise it
and txn->mt_toggle could become inconsistent, since read-only
txns fetched both from MDB_txbody without mutex proteciton.

Make DPRINTF usable in if-else.  Silence signed vs unsigned warning.

Also add mdb_debug/mdb_debug_start to toggle debug output at runtime

Was being fooled before because the newly malloc'd block had
the same address as the previously freed block.

Should probably compare nsize to pg_size/4 instead...