Launchable

Add validation to the launchable record build command https://github.com/launchableinc/cli/pull/292

Add inspect command to confirm subset result.
Usage: `launchable inspect subset --subset-id <subset id>`

* Status message improvements
* User agent format fix

Improve cli outputs like below

- launchable record tests

$ launchable record build --name cli
Launchable uploaded 8 commits from repository /launchableinc/cli
Launchable recorded build cli to workspace launchableinc/cli with commits from 1 repository

| Name | Path | HEAD Commit |
|--------|--------|------------------------------------------|
| . | . | 4729a34fffae4f6d6190288466e3826c5f3dd488 |


- launchable subset


$ find -f tests/**/test_*.py | launchable subset --build cli --target 50% file > subset.txt
Your model is currently in training
Launchable created subset 1379 for build cli (test session 305) in workspace launchableinc/cli

| | Candidates | Estimated duration (%) | Estimated duration (min) |
|-----------|--------------|--------------------------|----------------------------|
| Subset | 16 | 50 | 0.000266667 |
| Remainder | 16 | 50 | 0.000266667 |
| | | | |
| Total | 32 | 100 | 0.000533333 |


- launchable record tests


$ launchable record tests --build cli file ./test-results/*.xml
Launchable recorded tests for build cli (test session 306) to workspace launchableinc/cli from 30 files:

| File found | Tests found | Test passed | Test failed | Total duration(min) |
|--------------|---------------|---------------|---------------|-----------------------|
| 30 | 86 | 84 | 2 | 0.0601 |

- Automatically convert absolute file paths into relative file paths when communicating with the backend.

Previously you need to specify `--base` in order to make this conversion. From this version, if you do not specify the option, it automatically infers the base path and make file paths relative. You can suppress this inference with a CLI flag if needed.

- Add an option to skip commit data collection for the `record build` command.

Under a shallow clone repository, it's possible that a revision walk hits the depth limit, and the CLI errors out. With this new `--no-commit-collection` option, you can skip the commit data collection phase, and it just checks the HEAD commit hashes. You need to record the commit data separately in a fully cloned repository in order to get the training data right.

Fix the http client UserAgent include test runner name