#4932 - Fixes for field escaping in orquesta workflows

[nmaludy]

Closes #4932

This is a fix for the bug found in #4932 .

I had to revert 3x fields in the workflow DB model:

1. WorkflowExecutionDB.spec for any fields/dict keys that were statically defined in the workflow spec that include a . in their name:
   Example:

vars:
  - vars.key.with.periods: "vars.value.with.periods"
  - vars.nested.with.periods:
      nested.periods: "vars.nested.value.with.periods"

output:
  - wf.hostname.with.periods: "{{ ctx()['published.hosts'] }}"
  - wf.output.with.periods: "{{ ctx()['published.field'] }}"

2. WorkflowExecutionDB.context for any dicts/objects that is present in the "context", ie. from a published variable. This can occur, for example, when running core.remote. The resulting dictionary has hostnames/IPs as keys in the returned dict.
   Example:

  run:
    action: core.remote
    input:
      cmd: date
      hosts: "hostname.domain.tld,hostname2.domain.tld"
    next:
      - when: "{{ succeeded() }}"
        publish:
          - published.hosts: "{{ result().result }}"
          - published.field: "{{ result().result['hostname2.domain.tld']['stdout'] }}"

3. TaskExecutionDB.task_spec for any inputs passed into the task that may have a dict with a key contain a ..
   Example:

  bolt_plan_run:
    action: bolt.plan_run
    input:
      plan: "{{ ctx().plan }}"
      targets: "{{ ctx().targets }}"
      env:
        ST2KV_CONFIG:
          ssh.username: system.linux.username
          ssh.password: system.linux.password

I've added a single integration test case examples.orequest-test-field-escaping that invokes these three corner cases.

[m4dcoder]

LGTM

[winem]

LGTM!

And I successfully tested the fix on one of my dev nodes. I just added an additional variable some.value to the packs/actions/workflows/install.yaml

Without the fix:

With the fix applied:

[nmaludy]

FYI build failures were caused by poisoned Travis cache where pip==20.1 was installed and cached. Travis uses a hierarchical cache where if it can't find a cache for the current PR, it uses the cache from master. Since pip==20.1 breaks our fixate-requirements.py this caused both the master build and this PRs build to fail.

[arm4b]

This PR definitely needs a Changelog describing the bugfix.

[nmaludy]

@armab Added a changelog. Thank you for noticing this, i was in such a rush to get it fixed i forgot!

[arm4b]

👍 Thanks!