Support ingestion of tabular site forecasts and observations into EMOS

[gavinevans]

• Add dependencies on pandas and fastparquet #1572
• Add functionality to load a parquet file #1581
• Add functions to support conversion of a pandas DataFrame into an iris cube #1582
• Add CLI for ingesting tabular forecasts and observations into EMOS #1584 =>Replaced by Add CLI for ingesting tabular forecasts and observations into EMOS gavinevans/improver#16 => Replaced by Add CLI for ingesting tabular forecasts and observations into EMOS #1592

As an IMPROVER developer, I would like to support the ingestion of spot forecasts and observations into EMOS, so that a training dataset of spot forecasts and observations can be stored in a more usable format.

The forecast table provided will contain likely include percentiles in excess of those required for calibration. The estimate EMOS functionality therefore needs to:

1. Extract the required equally spaced percentiles from the required stream.
2. Rebadge the percentiles as realizations for compatibility for the existing EMOS codebase.

Acceptance criteria:

• Ability to read spot forecasts and observations in Parquet format into pandas dataframes.
• Support for use of pandas dataframes within the EMOS code.

[fionaRust]

All tasks now complete