Conflicts:
	libcelt/celt.c
	libcelt/celt.h

which allows the caller to request that frames be independent.

state before using it. Handle malloc failures for the encoder and
decoder setup gracefully and without leaks.

allocation

signal characteristics

allocator which was removed a while ago.

a really big amplitude change, in which case the original time-domain window
is used.

"tandem-test" which tests the encoder and decoder in
many modes.

the encoder could attempt to create folding-free short blocks which
is not permitted by the bitstream.

This avoids the need for 64-bit addition and is faster on x86_64.
It may be slower on some platforms so the direct solution is still
available in the source.

Adds specialized O(N*log(K)) versions of cwrsi() and O(N) versions of icwrs()
 for N={3,4,5}, which allows them to operate all the way up to the theoretical
 pulse limit without serious performance degredation.
Also substantially reduces the computation time and stack usage of
 get_required_bits().
On x86-64, this gives a 2% speed-up for 256 sample frames, and almost a 16%
 speed-up for 64 sample frames.

bitrates. It was caused by excessive "balance" bits leading to qb>14.

(rounding towards zero).

was removed from celt_mode_create()

wider range of values.

thus save a few divisions.

fixed point compilation because SCALEIN is a no-op in fixed point
mode but the float interface is still normally +/- 1.0.

This patch adds a seperate define for the scaling factor so people
can adjust it for unusual input levels.

that I'm resetting more than is necessary, but I would have to know a lot
more about how CELT works to determine that.