Reset delete flag in cursor_put().

If so, don't advance cursor on next NEXT. Also, init xcursor
on the next GET_CURRENT.

The caller is responsible for making sure no transactions are
active in this process before resizing. This is slightly lighter
weight than doing a full env_close/env_open cycle.

Use a txn instead of env argument. Only return persistent flags.

Always leave cursor pointing at "next" node, if any.
Find next sibling if we're already at end of current page.

If somehow "need" is larger than the list size, don't
try to fake out the dirty list.

spilling all possible pages is a waste because the majority will be
needed again. Just do 1/8th of the dirty list instead.

Do not access a long long as an array of longs.

For RDONLY, don't get lockfile until we know datafile exists.
Also, don't try to create a new datafile for me_mfd if someone
deleted it after we got me_fd.

This reverts commit 87258384.

Don't compare cursors beyond their depth. (detected by valgrind)

Should allow for the possibility of future flag combos.

If a key has a single existing value, and then a put (MDB_MULTIPLE)
is done where the first of the multiple values matches the existing
value, the put would return SUCCESS without writing any of the
values. Fixed to loop to the next value as intended.

Use double DPRINTF() parenthesis, so varargs macros are not needed.
Parenthesize MDB_DEBUG, in case the user defines it as foo || bar.

Also check xcursor instead of DUPSORT, it's slightly briefer.

Checking dirty_list was insufficient after a spill
with no named databases and no positioned cursors.

Return EINVAL only for simple programmer errors.

Ignore parent txn cursors since it is the current txn's dirty_list
which will be flushed. But check the current txn also when clearing,
since cursors can have pages which are dirty in a parent.

Check !mc_xcursor instead of !MDB_DUPSORT.  Equivalent for valid
data, but a bit safer if the sub-DB flags are corrupt.