Parallelize compression/decompression during backup and restore

[pliu]

Drop-in of pgzip for gzip

Compression for the backup and restore commands of influxd were able to download the data from the database quickly, but subsequent compression and decompression was done in a single-threaded manner that is extremely slow.

Required for all non-trivial PRs

• [ x ] Sign CLA (if not already signed)

[e-dard]

@pliu thanks for the contribution. Whilst the code changes in this PR are trivial, there may be significant changes to the behaviour of the database. In order to accept this contribution it would be good to see some benchmarks or performance tests.

Further, the core team would probably need to look into the third part library and check it's suitable for inclusion. /cc @aanthony1243

[pliu]

@e-dard Do you have internal benchmark/performance tests that you can run against the change or is this something I will need to perform?

Thanks!

[e-dard]

Hi @pliu this is looking fine. Thanks for this. We have recently started using dep as our dependency management tool. Do you mind if I add a commit to the PR with the correct dep update for pgzip?

[aanthony1243]

agreed, once we update for dep, this is ok to merge.

[pliu]

@e-dard Feel free to make the necessary changes :) Thank you!

[e-dard]

@pliu hi, can you cherry pick this commit into your branch? 4712404

[pliu]

@e-dard Done