Compress on the fly

[vchuravy]

Fixes #81

[NHDaly]

🤔 hrm tests are failing now, even though we are decoding. i'm not sure why. Any idea?

[NHDaly]

@adnan-alhomssi: This seems like a good idea, and is related to your suggestion that we should be compressing the profiles and heapsnapshots before sending them back from ProfileEndpoints.jl.

Can you have a look and see if you can help us figure out why the tests aren't passing?

[NHDaly]

Request for review: @Drvi

[codecov-commenter]

Codecov Report

Merging #82 (a59d315) into main (45399a8) will increase coverage by 0.06%. The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main      #82      +/-   ##
==========================================
+ Coverage   96.59%   96.66%   +0.06%     
==========================================
  Files           3        3              
  Lines         294      300       +6     
==========================================
+ Hits          284      290       +6     
  Misses         10       10              

Files	Coverage Δ
src/Allocs.jl	92.40% <100.00%> (+0.19%)	:arrow_up:
src/PProf.jl	99.19% <100.00%> (+0.01%)	:arrow_up:
src/flamegraphs.jl	96.90% <100.00%> (+0.06%)	:arrow_up:

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

[Drvi]

Will this break reading profiles from older PProf.jl versions? Typically with text files, one can sniff the gzip "magic number", but with ProtoBuf I think we need to check first if there are possible collisions with the encoded message.

[NHDaly]

Thankfully, i don't think PProf.jl ever reads profiles, we just hand them over to pprof, which seems to be able to read them fine, in either version.

So i think it's okay? What concern did you have in mind?

[Drvi]

Ah if we don't need to read it on Julia side, then it's fine:+1: sorry for the noise