Implement Sorted Writes

[vinjai]

Fixes: #271

[vinjai]

This PR solves for:

1. Writing sorted datasets to a partitioned or non-partitioned iceberg table.
2. Generating manifests with correct sort-order-id.
3. Integration tests to make sure sorted datasets are generated similar to spark sorting.

Decisions taken:

• If a sort transformation is not supported in PyIceberg, we will raise a warning related to the same and move ahead by writing the unsorted data with unsorted sort-order-id.

What is not in the scope of this PR?

• Performance improvement of the new sort function. (We will raise a separate issue for the same.)

[vinjai]

@Fokko This PR is ready for review

[vinjai]

Just wanted to clarify on the separate implementation for sort_indices other than the one provided by pyarrow.
This is because pyarrow sort_indices or Sort Options only supports one order for null placement across keys.
More details here:

• https://arrow.apache.org/docs/python/generated/pyarrow.compute.sort_indices.html
• https://arrow.apache.org/docs/python/generated/pyarrow.compute.SortOptions.html

While, the iceberg spec doesn't discriminate of having different null ordering across keys: https://iceberg.apache.org/spec/#sort-orders

This function specifically helps to implement the above functionality by sorting across keys and utilizing the stable nature of the sort_indices algo from pyarrow.

---

We can raise another issue to improve the performance of this function.

---

In future, if pyarrow sort_indices does support different null ordering across, we can mark this function as obsolete and keep the implementation clean in the iceberg table append and overwrite methods.