Handle incomplete/multiline NDJSON records in CVEListV5Handler and add tests

[adulau]

Motivation

• Improve robustness when ingesting NDJSON exports by correctly handling records that are split across lines or end up as incomplete JSON fragments so subsequent records are not lost.

Description

• Update process_ndjson_file to buffer consecutive lines into pending_lines and attempt to parse a combined payload before skipping.
• Add helper methods �_load_record�, �_is_incomplete_record�, �_looks_like_record_start�, �_process_record�, and �_skip_invalid_record� and the INCOMPLETE_ERROR_PREFIXES constant to centralize parsing and error handling.
• Preserve existing behavior of recording skipped records and progress via record_progress and process_rank_batch while skipping only truly invalid or incomplete records.
• Add unit test test_skips_single_invalid_multiline_record_and_continues to verify a broken multiline record is skipped and processing continues for following valid records.

Testing

• Ran the tests/test_cvelistv5.py unit tests which include the new test_skips_single_invalid_multiline_record_and_continues test.
• All unit tests passed locally with the updated handler and test additions.

---

Codex Task