Make extracted raw data location configurable

[alistairsellar]

The standardised output, and also raw extracted data from standardise_model_data is currently being written to share/data, see app/standardise_model_data/rose-app.conf. The output from standardise_model_data should continue to be written to share/data since it is used by the run_recipe_* tasks, but the extracted raw data is only used by this task and should be written to work.

Include configurable extract location to support subsequent runs that switch off the extract step.

Acceptance criteria 1:

• Given configuration details of a location outside the cylc-run directory to extract raw data to: RAW_DATA_DIR
• When standardise_model_data runs
• Then the raw model data is written to the extract dir

Acceptance criteria 2:

• Given configuration details in which RAW_DATA_DIR is not set,
• When standardise_model_data runs
• Then the raw model data is written to a dir under share/data within the cylc-run directory (as at present)

Details

• See also Develop a housekeeping task #30 (though they are not interdependent).
• It would be better to work on this after Consolidate ROOT_PROC_DIR and ROOT_DATA_DIR #151 / Consolidate ROOT_PROC_DIR and ROOT_DATA_DIR #205 are complete.
• Estimate: medium.

[mo-nabilbrice]

From PR #172, there is a need to point the output directory from restructure_for_cmip6, which is set in the bash script using the ROOT_RESTRUCTURED_DIR variable, to be in the work directory rather than the share/etc directory.

[alistairsellar]

The Cylc documentation indicates that data which is used by multiple tasks should be written to share/data: https://cylc.github.io/cylc-doc/8.2.5/html/workflow-design-guide/general-principles.html#shared-task-io-paths

I think therefore that we can use work for the raw extracted model data, which is used only by the standardise_model_data, but that the standardised data should continue to use share/data because it is required by the run_recipe_* tasks.