feat: add CSV-based upload support and patch option for existing records

Problem Statement

Currently, the corpus-client upload_files command uploads files from a single source (local directory or S3 bucket) with shared metadata entered via CLI prompts. Users need to manually enter fields like title, description, language, etc., for every batch upload, even when uploading multiple files with different metadata. Additionally, there's no way to update existing records in bulk using CSV data.

Feature 1: CSV-based Upload for New Records

Goal

Enable batch uploads where each file can have different metadata via CSV input, eliminating manual CLI prompts.

Proposed CLI Usage

corpus-client upload_files /path/to/files --csv /path/to/metadata.csv

Expected CSV Format

filename	title	description	language	category_ids	release_rights	creator	latitude	longitude	source_label	published_date
audio1.mp3	Folk Song from Kerala	A beautiful traditional...	malayalam	cat-uuid-1,cat-uuid-2	creator		10.8505	76.2711	SOAI 2025	2024-01-15

Required columns

• filename - relative path to the file in the provided directory
• title
• description (min 32 chars)
• language
• category_ids (comma-separated UUIDs)
• release_rights

Optional columns

• creator (required if release_rights=others)
• latitude
• longitude
• source_label
• published_date

Behavior

• If CSV is provided, fail if any required row is missing its value
• Files are resolved relative to the provided path argument (local dir or S3 prefix)
• Auto-detect media_type from file extension if not provided

Feature 2: --patch Flag for Updating Existing Records

Goal

Update existing records using CSV data, where each row contains a record uid and the fields to modify.

Proposed CLI Usage

corpus-client upload_files /path/to/files --csv /path/to/updates.csv --patch

Expected CSV Format

uid	title	description	language	release_rights	creator	latitude	longitude	source_label
record-uid-1	Updated Title	Updated description				17.4459	78.3504
record-uid-2			kannada	others	Author Name

Behavior

• Only non-empty fields in CSV are updated (PATCH semantics)
• Empty cells = no change to that field
• uid is required for identifying which record to update
• Calls backend PATCH /api/v1/records/{record_id} endpoint

PATCH-able fields

• title
• description
• location (latitude/longitude)
• release_rights
• language
• published_date
• creator
• source_label
• source_url

Backward Compatibility

• If CSV is NOT provided, fall back to current interactive prompt flow
• Work with single file, directory, or S3 URI as before

Additional Considerations

• CSV parsing should handle comma-separated category_ids (e.g., "cat-id-1,cat-id-2")
• Validate CSV headers before processing
• Show preview of records to be uploaded/updated before proceeding
• Support resume for interrupted CSV batch operations