feat(float): upgrade SDK to 2.0.0, add pytest unit tests (#284)

* feat(float): upgrade SDK to 2.0.0, add pytest unit tests

- Bump autohive-integrations-sdk pin from ==1.0.2 to ~=2.0.0
- Update all context.fetch() call sites to access .data on FetchResponse
- Fix connected account handler to use response.data for isinstance check
- Bump config.json version to 2.0.0
- Create float/tests/conftest.py with standard sys.path insert
- Create float/tests/test_float_unit.py with 71 unit tests covering
  people, projects, tasks, time off, logged time, clients, departments,
  roles, reference data, holidays, phases, milestones, reports, and
  auth header verification

* fix(float): apply Kai SDK skills — fix tests, FetchResponse, ResultType assertions

- Add ActionError import and convert all raise Exception() error paths to return ActionError(message=...)
- Fix action handlers to use response.data from FetchResponse
- Fix connected account handler to use response.data
- Replace integration-style test_float.py with proper unit tests split by domain
- 106 unit tests across 5 files: people, projects, tasks, timeoff, misc
- All tests use FetchResponse mocks and ResultType.ACTION_ERROR assertions per skill

* fix(float): fix 3 error path tests to use ResultType.ACTION_ERROR instead of pytest.raises

* fix(float): fix auth, optional params, add proper unit and integration tests

* fix(float): bump SDK to 2.0.0, format tests, verify 5/5 live integration tests passing

* fix(float): bump version to 2.0.0, fix isinstance to use response.data

* style(float): apply ruff formatting

* fix(float): remove os.chdir, delete broken legacy test_float.py and context.py

* test(float): expand integration tests to cover all 52 actions

* fix(float): fix config.json output schemas and integration test assertions

- Make all output schema field types nullable to match Float API reality
  (API returns null for many optional fields that were typed as non-nullable)
- Change `active` field type from boolean to [integer, boolean, null] since
  Float API returns 1/0 integers, not booleans
- Fix list_statuses to return [] instead of None on 204 No Content response
- Fix integration test assertions:
  - List endpoints: assert isinstance(data, list) not wrapped key access
  - Get endpoints: assert isinstance(data, dict) not wrapped key access
  - Use task_meta_id fallback for project_task items (API uses different key)
  - Use id fallback for public_holiday and project_stage items
  - Destructive tests: access fields directly on dict not via nested wrapper

* style(float): apply ruff formatting to test_float_integration.py

---------

Co-authored-by: Shubhank <72601061+Sagsgit@users.noreply.github.com>
Co-authored-by: Kai Koenig <kai@ventego-creative.co.nz>

Author: Shubhank-Jonnada

.env.example

@@ -60,6 +60,12 @@
 # -- Xero --
 # (uses platform OAuth — tokens are short-lived, typically not set here)
 
+# -- Gong --
+# GONG_ACCESS_KEY=
+# GONG_ACCESS_KEY_SECRET=
+
+# -- Float --
+# FLOAT_API_KEY=
 # -- Supadata --
 # SUPADATA_API_KEY=
 # -- LinkedIn --