fix: Fix the panic during parallel unit mapping conversion.

[shanicky]

I hereby agree to the terms of the RisingWave Labs, Inc. Contributor License Agreement.

What's changed and what's your intention?

Title:

Refactor consistent hashing and improve error handling in metadata controllers

Summary:

This pull request introduces a series of updates to Rust files associated with consistent hashing and metadata controllers in our distributed system. Below is an overview of the main changes incorporated into the system:

Changes to consistent_hash/masking.rs:

• Introduced a new HashSet import from std::collections
• Added a new Sub import from std::ops
• Implemented a new error type ParallelUnitError to handle un-mapped parallel units in the system
• Updated ParallelUnitMapping::to_worker_slot to return a Result type, now checking complete coverage of parallel unit IDs by worker slots and returning an error if uncovered
• Added method as_delete_worker_slap_mapping to ParallelUnitMapping for creating a WorkerSlotMapping indicating delete operations via max u64 values

Modifications to catalog.rs:

• Removed an unused ParallelUnitMapping import
• Altered methods (drop_database, deploy_stream_job, stop_stream_job) to set the worker slot mapping to None in PbFragmentWorkerSlotMapping instead of computing a mapping indicating a shift in mapping handling
• Extricated the obsolete get_parallel_unit_to_worker_map function from several methods

Updates to fragment.rs and streaming_job.rs:

• Adapted both files to account for ParallelUnitMapping::to_worker_slot now returning a Result, enabling proper error propagation
• Created a helper function convert_fragment_mappings in fragment',' which converts fragment mappings to protobuf representations while handling errors from to_worker_slot`

The overarching goal of these changes is to streamline and enhance error handling for converting parallel unit mappings to worker slot mappings, which is a pivotal aspect of our metadata controllers. Affected code segments now anticipate potential conversion errors and are equipped for their management, markedly improving system robustness in handling parallel unit mapping scenarios.

Note: Reviewers should carefully assess the potential impacts of setting the worker slot mappings to None and confirm that the propagation of conversion errors aligns with the intended behavior of the distributed system architecture.

Checklist

• I have written necessary rustdoc comments
• I have added necessary unit tests and integration tests
• All checks passed in ./risedev check (or alias, ./risedev c)