Mara-pipelines

- Changed all `TIMSTAMP` to `TIMSTAMPTZ` in the mara tables. You have to manually run the
below migration commands as `make migrate-mara-db` won't pick up this change.

**required changes**
You need to manually convert the mara tables to `TIMESTAMPTZ`:

SQL
-- Change the timezone to whatever your ETL process is running in
ALTER TABLE data_integration_run ALTER start_time TYPE timestamptz
USING start_time AT TIME ZONE 'Europe/Berlin';
ALTER TABLE data_integration_run ALTER end_time TYPE timestamptz
USING end_time AT TIME ZONE 'Europe/Berlin';
ALTER TABLE data_integration_processed_file ALTER last_modified_timestamp TYPE timestamptz
USING last_modified_timestamp AT TIME ZONE 'Europe/Berlin';
ALTER TABLE data_integration_node_run ALTER start_time TYPE timestamptz
USING start_time AT TIME ZONE 'Europe/Berlin';
ALTER TABLE data_integration_node_run ALTER end_time TYPE timestamptz
USING end_time AT TIME ZONE 'Europe/Berlin';
ALTER TABLE data_integration_node_output ALTER timestamp TYPE timestamptz
USING timestamp AT TIME ZONE 'Europe/Berlin';
ALTER TABLE data_integration_file_dependency ALTER timestamp TYPE timestamptz
USING timestamp AT TIME ZONE 'Europe/Berlin';
ALTER TABLE data_integration_system_statistics ALTER timestamp TYPE timestamptz
USING timestamp AT TIME ZONE 'Europe/Berlin';

- Track and visualize also unfinished pipeline runs
- Speed up computation of node durations and node cost
- Improve error handling in launching of parallel tasks
- Improve run times visualization (better axis labels, independent tooltips)
- Smaller ui improvements

- Remove dependency_links from setup.py to regain compatibility with recent pip versions
- Change MARA_XXX variables to functions to delay importing of imports
- move some imports into the functions that use them in order to improve loading speed
- Add ability to mask passwords in `Command`s, so they cannot show up in the UI anymore
or are not written to the database in saved Events (config
`data_integration.config.password_masks()`). See the example in the original function
how to not let passwords show up in the settings UI.
([gh 14](https://github.com/mara/data-integration/pull/14))

**required changes**

- Update `mara-app` to `>=2.0.0`

- Use postgresql 10 native partitioning for creating day_id partitions in ParallelReadFile
- Catch and display exceptions when creating html command documentation
- Add python ParallelRunFunction
- Add option to use explicit upsert on incremental load (explicit UPDATE + INSERT)
- Emit a proper NodeFinished event when the launching of a parallel task failed
- Add option truncate_partition to parallel tasks
- Fix bug in run_interactively cli command
- Make it possible to run the ExecuteSQL command outside of a pipeline via .run()
- Add args parameter to RunFunction command
- Show redundant node upstreams as dashed line in pipeline graphs
- Fix problems with too long bash commands by using multiple commands for partition generation in ParallelReadXXX tasks

**required changes**

- When using `ParallelReadFile` with parameter `partition_target_table_by_day_id=True`, then make sure the target table is natively partitioned by adding `PARTITION BY LIST (day_id)`.

- Add possibility to continue running child nodes on error (new `Pipeline` parameters `continue_on_error` and `force_run_all_children`)
- Make dependency on requests explicit

- Implement ReadMode ONLY_CHANGED that reads all new or modified files
- Show node links in run output only relative to current node (to save space)