Fixing How2-2000h Data preparation and Seq Length Assert for Longformer Encoder

[roshansh-cmu]

We changed the data download mechanism for How2-2000h recently. This impacts the asr1 and sum1 recipes for how2_2000h.
Therefore, we updated the README and data preparation scripts.

• Data Prep and README for asr1
• Data Prep and README for sum1
• Sequence length assert should not be checked in espnet_model.py for longformer since we may do some padding in the encoder to optimize GPU usage. This is a package default and not easy to modify at this time, so the assert runs always except when the encoder's selfattn is "lf_selfattn".
• Add HOW2_2kH to db.sh

Please consider and merge at your convenience.

[codecov]

Codecov Report

Merging #4805 (01dbafa) into master (1a50788) will increase coverage by 0.00%.
The diff coverage is 100.00%.

@@           Coverage Diff           @@
##           master    #4805   +/-   ##
=======================================
  Coverage   80.38%   80.38%           
=======================================
  Files         533      533           
  Lines       46948    46949    +1     
=======================================
+ Hits        37738    37739    +1     
  Misses       9210     9210           

Flag	Coverage Δ
test_integration_espnet1	66.37% <ø> (ø)
test_integration_espnet2	49.00% <100.00%> (-0.02%)	⬇️
test_python	68.75% <100.00%> (+<0.01%)	⬆️
test_utils	23.30% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
espnet2/asr/espnet_model.py	81.06% <100.00%> (+0.07%)	⬆️

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[sw005320]

Thanks a lot, @roshansh-cmu!
After the CI check, I'll merge this PR.