accept memory and gpu memory in GiB instead of MiB

[bwagner5]

Issue #, if available:
N/A

Description of changes:

• Accept GiB instead of MiB to memory and gpu-memory

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

[brycahta]

Looks good overall! just some nits with comment corrections.

Out of curiosity, did you consider taking either as input (MiB or GiB) in an effort to maintain backwards compat? I feel like one could establish some reasonable bounds to determine whether the input was MiB or GiB, but figured you'd have more context/already figured out it wouldn't work.

[codecov-commenter]

Codecov Report

Merging #36 into master will decrease coverage by 1.13%.
The diff coverage is 81.32%.

@@            Coverage Diff             @@
##           master      #36      +/-   ##
==========================================
- Coverage   90.05%   88.92%   -1.14%     
==========================================
  Files           8        9       +1     
  Lines         875     1056     +181     
==========================================
+ Hits          788      939     +151     
- Misses         57       81      +24     
- Partials       30       36       +6     

Impacted Files	Coverage Δ
pkg/selector/types.go	100.00% <ø> (ø)
pkg/selector/selector.go	84.51% <31.25%> (-0.80%)	⬇️
pkg/cli/types.go	95.28% <78.26%> (-4.72%)	⬇️
pkg/cli/cli.go	86.00% <82.35%> (+0.51%)	⬆️
pkg/cli/flags.go	87.70% <83.33%> (-2.82%)	⬇️
pkg/bytequantity/bytequantity.go	87.30% <87.30%> (ø)
pkg/selector/aggregates.go	92.45% <100.00%> (ø)
pkg/selector/comparators.go	91.46% <100.00%> (+0.43%)	⬆️
pkg/selector/outputs/outputs.go	91.81% <100.00%> (ø)
... and 1 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update be67b3f...5dd8215. Read the comment docs.

[bwagner5]

@brycahta I don't think it would work to accept both without a suffix. There are currently 32 instance types that have between 512-24,576 GiB of memory and 88 instance types that have between 512-21,000 MiB which would overlap the top 32. The user experience of not knowing exactly what the input will be parsed as seems bad too.

I did consider a suffix --memory 512mb or --memory 512gb which I think makes sense. But it still seems like it was a mistake to default to MiB, so I still think we need to change to GiB as the default, and probably better earlier rather than later.

--memory 512 == --memory 512gb
--memory 0.5 == --memory 512mb

[brycahta]

add suffixes

changes

[bwagner5]

there seems to be something wrong with this travis VM. I ran the tests on the same branch on my fork and everything passes: https://travis-ci.org/github/bwagner5/amazon-ec2-instance-selector/builds/711445177

[brycahta]

had a couple minor comments but overall looks solid 👍🏼 Loving the support for all the units!

[brycahta]

LGTM👍🏼

[haugenj]

LGTM 👍