experiments: support column ordering & sorting in dvc exp show

[pmrowla]

• ❗ I have followed the Contributing to DVC checklist.

• 📖 If this PR requires documentation updates, I have created a separate PR (or issue, at least) in dvc.org and linked it here.

Thank you for the contribution - we'll try to review it as soon as possible. 🙏

Will close #4346.

• When using --include-metrics/--include-params, columns will be displayed left-to-right in the order specified by the --include option.
• --sort-by <column_name> and --sort-order asc|desc can be used to sort related experiments on the specified metric or param column.
  • Only nested experiments will be sorted, baseline (formal) commits will still be displayed in the git history order.
  • --sort-order defaults to ascending
  • Only sorting by a single column is supported

❯ dvc exp show --include-params=train.seed,featurize --sort-by=featurize.max_features --sort-order desc --all-commits --no-pager
┏━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━┳━━━━━━━━━┳━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┓
┃ Experiment            ┃ Created      ┃     auc ┃ train.seed ┃ featurize.max_features ┃ featurize.ngrams ┃
┡━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━╇━━━━━━━━━╇━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━┩
│ workspace             │ -            │ 0.58164 │ 20170428   │ 1500                   │ 3                │
│ executor-tree         │ Aug 28, 2020 │ 0.50822 │ 20170428   │ 1500                   │ 4                │
│ └── *41a9f4b          │ Aug 28, 2020 │       - │ 20170428   │ 456                    │ 5                │
│ 6d18fc1               │ Aug 24, 2020 │ 0.58164 │ 20170428   │ 1500                   │ 3                │
│ ├── 5840202           │ Aug 24, 2020 │ 0.50822 │ 20170428   │ 1500                   │ 4                │
│ ├── 3c8db11           │ Aug 24, 2020 │ 0.50298 │ 20170428   │ 500                    │ 4                │
│ └── *3b6e442          │ Aug 25, 2020 │       - │ 20170428   │ 123                    │ 5                │
│ 0562956               │ Aug 24, 2020 │ 0.61314 │ 20170428   │ 1500                   │ 2                │
│ ├── e6d1716           │ Aug 24, 2020 │ 0.58164 │ 20170428   │ 1500                   │ 3                │
│ └── *c34fd7e          │ Aug 25, 2020 │       - │ 20170428   │ 456                    │ 5                │
│ acd59eb               │ Jul 22, 2020 │ 0.61314 │ 20170428   │ 1500                   │ 2                │
...
│ 0-git-init            │ Jun 20, 2020 │       - │ 20170428   │ 1500                   │ 3                │
└───────────────────────┴──────────────┴─────────┴────────────┴────────────────────────┴──────────────────┘

[pmrowla]

Use OrderedDict() with None values to get set-like behavior with preserved ordering

[skshetry]

Or, use a dict?

[pmrowla]

I thought the preserved key ordering for regular dict technically isn't guaranteed before 3.7 (it's a cpython implementation behavior in 3.6 that was added to the formal spec in 3.7), so I thought it would be preferred to just use OrderedDict?

[skshetry]

yes, you are right that it is only on >3.6. My point was that, as CPython and Python are closely tied, no one is going to change that behavior in 3.6, so should be safe to just use dict for convenience. Saying that I have no issues with OrderedDict, it was just a suggestion.

[skshetry]

Looks good. Just one suggestion.

[skshetry]

Or, use a dict?