Please, review @Tyagov, @kirr and @tatuya. 

All the information of this merge request is in the commit message, I'm pasting here for convenience. 

## Pandas-based Inventory API Protoype

The implementation relies on the Data Array Module. It imports data from the
stocks table through a zSQL Method. Category information is added later in a
column-wise way, so it can be easily done in parallel and query Portal Catalog
once for each category column in the array. This category processing needs to be
done only once, when the array is created, and to new data as it is added.

But there is a catch: each entity that belongs to the movement can have many
categories. So the row can be duplicated for each entity's categories and
searched by equality, or they can be stored as comma-separated values and
searched using a regular expression. Regular expression seems faster for
datasets up to 1M rows.

Some unit tests were also added.

These are the external methods created and their purposes:

- Base_filterInventoryDataFrame is there just to parse keyword arguments and
forward them to Base_getInventoryDataFrame. This is used for the non-programmer
interface of Pandas-based getMovementHistoryList implementation and can be
used as an external method in other scripts too.

- Base_convertResultsToBigArray will convert results of Portal Catalog and ZSQL
Method to a Data Array with a proer transformation of the schema to a
compatible NumPy data type.

- Base_extendBigArray will extend a Data Array with a Portal Catalog query or
ZSQL Method result. Raise errors when the extension data type doesn't match
the source.

- Base_fillPandasInventoryCategoryList will fill category infomration in a Data
Array which has stock movements information.

- Base_zGetStockByResource is used in a test case as source to create a Data
Array with stock data. 

/reviewed-on nexedi/wendelin!10

The implementation relies on the Data Array Module. It imports data from the
stocks table through a zSQL Method. Category information is added later in a
column-wise way, so it can be easily done in parallel and query Portal Catalog
once for each category column in the array. This category processing needs to be
done only once, when the array is created, and to new data as it is added.

But there is a catch: each entity that belongs to the movement can have many
categories. So the row can be duplicated for each entity's categories and
searched by equality, or they can be stored as comma-separated values and
searched using a regular expression. Regular expression seems faster for
datasets up to 1M rows.

Some unit tests were also added.

These are the external methods created and their purposes:

- Base_filterInventoryDataFrame is there just to parse keyword arguments and
forward them to Base_getInventoryDataFrame. This is used for the non-programmer
interface of Pandas-based getMovementHistoryList implementation and can be
used as an external method in other scripts too.

- Base_convertResultsToBigArray will convert results of Portal Catalog and ZSQL
Method to a Data Array with a proer transformation of the schema to a
compatible NumPy data type.

- Base_extendBigArray will extend a Data Array with a Portal Catalog query or
ZSQL Method result. Raise errors when the extension data type doesn't match
the source.

- Base_fillPandasInventoryCategoryList will fill category information in a Data
Array which has stock movements information.

-Base_zGetStockByResource is used in a test case as source to create a Data
Array with stock data.

The Append File control is hidden when the Data Stream is empty and enabled when
the Data Stream already has content. Meaninful notes were added to both Upload
File and Append File control to avoid confusion.

This business template was migrated to the new format during this
implementation.

/reviewed-on nexedi/wendelin!11

The Append File control is hidden when the Data Stream is empty and enabled
when the Data Stream already has content. Meaninful notes were added to both
Upload File and Append File control to avoid confusion.
This business template was migrated to the new format during this
implementation.

subset of bigger one.

Master

@Tyagov 
This Merge request adds
 * Data Array Line -> to get a view into data array in any dimension defined by numpy indexing syntax
 * Data Event Module -> to store user-entered information about monitoring, for example on missing data

The merge request also fixes getSize on empty data array and it fixes http range requests for some arrays.

See merge request !9

reason: previous way did not work for all kind of arrays

To avoid random test failures due to test execution we make start date one day before. This commit only fixes test.

Add comments to do.

This script should only append data to stream and NOT do any transformations on it keeping a simple rule of ERP5: save exactly what was entered by user or passed to us by remote agent (fluentd).

add array preview listbox to Data Array View

The listbox shows lines for all indexes in the first dimension of the ndarray and up to 100 columns for the second dimension.

See merge request !8

The listbox shows lines for all indexes in the first dimension of the ndarray and up to 100 columns for the second dimension.

Call pure transformation in an activity rather than execute in current transaction. This way we split ingestion part from transformation part.
Note: this commit will serialize argument_list to activity's mysql table in case of big packets this can be slow, in case of small appends it can be acceptable.
Instead we should call with start and end offset only and data should be read from Data Stream itself (WIP).

Raise in case of improper server side configuration so sender is aware rather than silently return nothing and fool sender it's all ok.
Pylint.