Initial support for Hive Catalog.

[subkanthi]

closes: #616

[subkanthi]

The current implementation of commitTable is not atomic, using Hive Locking API similar to iceberg-python

        # https://github.com/apache/hive/blob/master/standalone-metastore/metastore-common/src/main/thrift/hive_metastore.thrift#L1232
        with self._client as open_client:
            lock: LockResponse = open_client.lock(self._create_lock_request(database_name, table_name))

            try:
                if lock.state != LockState.ACQUIRED:
                    if lock.state == LockState.WAITING:
                        self._wait_for_lock(database_name, table_name, lock.lockid, open_client)
                    else:
                        raise CommitFailedException(f"Failed to acquire lock for {table_identifier}, state: {lock.state}")

[zeroshade]

Sorry for the long delay here, it's been a very busy couple weeks. Here's my first pass, more comments to come later tonight/tomorrow

[zeroshade]

Not using an io.Closer where the Close method returns error?

[zeroshade]

is there any particular reason to export this? Is the intent that users should be able to provide their own HiveClient if they want?

[zeroshade]

Do we expect users to actually be using this directly as opposed to just using NewCatalog?

[zeroshade]

If I'm reading gohive correctly, this requires the kerberos tag to build correctly, right? Can we instead use one of the pure go implementations that don't rely on cgo? Like gokrb etc.

[zeroshade]

Shouldn't we set c.client = nil after we close?

[zeroshade]

Most of the methods here just forward to client, should we just embed the client rather than have it be a member?