Fix desync with aarch64 assembly for ipred_z2

[barrbrain]

Parameters max_w and max_h were inadvertently swapped.

Before this patch:

$ rav1e -o test.ivf -r rec.y4m --limit 1 tractor_1080p10_25.y4m -y
>  CPU Feature Level: NEON
>  Using y4m decoder: 1920x1080p @ 25/1 fps, 4:2:0, 10-bit
>  Encoding settings: keyint_min=12 keyint_max=240 quantizer=100 bitrate=0 min_quantizer=0 low_latency=false tune=Psychovisual rdo_lookahead_frames=20 multiref=true fast_deblock=false scene_detection_mode=Standard cdef=true lrf=true enable_timing_info=false min_block_size=8x8 max_block_size=64x64 encode_bottomup=false non_square_partition_max_threshold=8x8 reduced_tx_set=true tx_domain_distortion=false tx_domain_rate=false rdo_tx_decision=false prediction_modes=Complex-KFs fine_directional_intra=true include_near_mvs=false use_satd_subpel=true
>  Using 1 tile
>  encoded 1/1 frames, 0.045 fps, 38630.00 Kb/s, est: 0s, 0.18 MB, elapsed: 22s                 
>  ----------
>  Key frame:             1 | avg QP:  77.00 | avg size:  193150 B
>  Inter frame:           0 | avg QP:   0.00 | avg size:       0 B
>  Intra only frame:      0 | avg QP:   0.00 | avg size:       0 B
>  Switching frame:       0 | avg QP:   0.00 | avg size:       0 B
>  ----

$ dav1d -i test.ivf -o ref.y4m
dav1d 1.3.0 - by VideoLAN
Decoded 1/1 frames (100.0%) - 33.88 fps

$ cmp <(tail -n+2 ref.y4m) <(tail -n+2 rec.y4m)
/dev/fd/63 /dev/fd/62 differ: byte 4075905, line 7294

After this patch cmp returns 0, i.e. the frames match perfectly.

Input tractor_1080p10_25.y4m was created from tractor_1080p25.y4m at https://media.xiph.org/video/derf/ :

ffmpeg -i https://media.xiph.org/video/derf/y4m/tractor_1080p25.y4m \
  -vframes 1 -pix_fmt yuv420p10le -strict -2 tractor_1080p10_25.y4m

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (162a6de) 88.22% compared to head (0fb3e3f) 88.22%.

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #3299   +/-   ##
=======================================
  Coverage   88.22%   88.22%           
=======================================
  Files          87       87           
  Lines       28221    28221           
=======================================
  Hits        24898    24898           
  Misses       3323     3323           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[barrbrain]

Credit to @vibhoothi for running the encoder on aarch64 with a diverse corpus.