Fixes an assert in dav1d_prepare_intra_edges_16bpc() with
clusterfuzz-testcase-minimized-dav1d_fuzzer_mt-5640098964570112. Credits
to oss-fuzz.

Adaption of the avx2 code

Fixes NULL dereference with
clusterfuzz-testcase-minimized-dav1d_fuzzer_mt-5649526686220288. Credits
to oss-fuzz.

Fixes green lines in #197.

Fixes #196.

Fixes subsequent decoding errors after unrelated decoding errors or
flush/seek events.

Based on the implementation from tools/dav1d.c

Any benchmark that uses random data as input gives bunk results as it
 currently uses differently random data on each run.
This now makes any non-determinism in the tests repeatable across each
 call to check_cpu_flags() and checkasm_check_func().

Fixes use of uninitialized values in
clusterfuzz-testcase-minimized-dav1d_fuzzer_mt-5767052292259840. Credits
to oss-fuzz.

Fixes a NULL derefence with
clusterfuzz-testcase-minimized-dav1d_fuzzer-5724731362246656. Credits to
oss-fuzz.

Otherwise the library user will have to guess and find out the limit the hard way.

Should increase function coverage in oss-fuzz to 100%.

Signed-off-by: James Almer <jamrial@gmail.com>

This partially reverts commit b6bb8536.

Relegating the work of deriving this value to the library user from
ss_hor, ss_ver and monochrome in functions that don't return a
Dav1dPictureParameters is not a good idea.

Signed-off-by: James Almer <jamrial@gmail.com>

The relative speedup compared to C code is around 4.2 for a Cortex A53
and 5.1 for a Snapdragon 835 (compared to GCC's autovectorized code),
6-7x compared to GCC's output without autovectorization, and ~8x
compared to clang's output (which doesn't seem to try to vectorize
this function).

This means storing monochrome, ss_hor/ver and hbd. The derived values
bpc and layout are then stored in Dav1dPictureParameters.

Also remove redundant entries from Dav1dPictureParameters, and move
documentation of these fields into Dav1dFrame/SequenceHeader instead.

Add DAV1D_/Dav1d prefix to everything.