Measured perf cost of meminit was 7% worst case on an Intel Core2
system. YMMV.

Silence warnings, fix broken MinGW64 definitions, etc.

Init malloc'd memory by default, use #MDB_NOMEMINIT to disable.
Bump version to 0.9.10

Initialize unused portions of DB pages, for security or to silence
checkers like Purify and Valgrind. Like BDB configure --enable-umrw.

A node may need room for [key, subpage with 2 items] = 3 keys.

Use DB page size = min(32k, OS pagesize).  Previous limit was 8k
(MDB_MINKEYS*MDB_PAGESIZE).

Handle DB pagesize < OS pagesize. That's an I/O pessimization,
but transactions remain atomic: Only writing the MDB_meta must
be atomic, and it fits in one OS page.

Don't truncate desired subpage size: Asssign it to a size_t
(mv_size), not an uint16_t (mp_upper).

Skip cursors that aren't as deep as current cursor

Tweak split_indx check, go one slot further.

In d8eccb35

Return the environment's filedescriptor. Useful when the caller
is doing their own locking.

USE_POSIX_SEM must define USE_HASH

Also, check the split point on branch pages as well as leaf pages.

There may be other uses for it besides semaphore names.

Other cursors being fixed up are not necessarily height 1.

Don't allow shrinking below minimum size for an already open env.

Do not fail after mdb_page_new() succeeds.

Both were unused and md_name was unmaintained -- except
mdb_cursor_touch(xcursor) would abuse md_name as a key to
touch MAIN_DBI if it could somehow get passed ! DB_DIRTY.

xcursor DBIs were parent DBI+1 for debugging. Instead output
-(parent DBI).  Fixes a crash in mdb_cursor_del0()'s xcursor
tracking, it forgot to subtract 1 for C_SUB cursors.

Leave all lock management to the caller.

Calling app wants to manage its own locking.