add an arrow::compute::ExecContext object to the cylon::context

[nirandaperera]

since we are using a lot of arrow compute functions, it makes sense to add an arrow::compute::ExecContext object to the cylon::context . Arrow compute ops, always accept an exec context, and the default value is nullptr.

[nirandaperera]

@chathurawidanage @supunkamburugamuve I think this is important. I am not sure how we could support this ATM. My initial idea was to add it to the cylon::context, but it does not have any arrow dependencies. What should be a good design here? may be extending cylon::context?

[chathurawidanage]

Can't we handle that similar to how we have handled MemoryPools?