Create Pybind wrappers for the Quantiles class

[dvadym]

Create Pybind wrappers for the Quantiles class

Create Pybind11 wrappers for the differential_privacy::Quantiles. This will expose the functionality to the PyDP library.

Background:

The C++ Differential Privacy library includes a Quantiles class for calculating differentially private quantiles ( source). The PyDP library, which provides Python wrappers for the C++ library, already has bindings for the QuantileTree class ( Python differential pr...). To provide a more complete and user-friendly API, we should also create wrappers for the higher-level Quantiles class.

Implementation Details:

The new wrappers should be created in a new file, pydp/src/bindings/PyDP/algorithms/quantiles.cpp. The Python API should follow the existing conventions in PyDP and mirror the C++ API of the Quantiles class as closely as possible.

The Quantiles class is templated on arithmetic types (int, double, etc.). The Pybind wrappers should be instantiated for common numerical only for double.

Classes and Methods to Wrap:

1. Quantiles.__init__: which calls Quantiles<double>::Builder. It should have all parameters as in Builder.

2. Quantiles

The following public methods of the Quantiles class should be wrapped:

AddEntry(const T&)
Serialize() const
Merge(const Summary&)
GetQuantiles() const
Result()

Considerations:

1. The Build() method returns an absl::StatusOr. The wrappers should handle the status and raise an exception on failure.

2. The SetLaplaceMechanism method takes a std::unique_ptr. For Python binding let's use enum - Laplace, Gaussian.