fix(buffers): treat I/O errors as recoverable to prevent panics

[huseynsnmz]

Summary

This PR fixes a panic in the disk buffer that crashes the entire Vector process when encountering I/O errors (such as missing buffer files).

The Problem:
When the disk buffer reader encounters I/O errors like NotFound (file missing), it panics at receiver.rs:59 with:

thread 'vector-worker' panicked at receiver.rs:59:33:
Reader encountered unrecoverable error: Io { source: Os { code: 2, kind: NotFound, message: "No such file or directory" } }

This panic crashes Vector completely, stopping all pipelines across all components.

The Solution:
Treat I/O errors the same way we already treat corruption/checksum errors - emit an error event, log it, update metrics, and skip to the next buffer file. The disk buffer reader already has roll_to_next_data_file() logic that handles advancing past problematic files, so we simply remove the panic and emit a BufferReadError instead.

Behavior Change:

• Before: I/O error → panic → entire Vector process crashes → manual restart required
• After: I/O error → log ERROR + emit metric → skip corrupted file → pipeline continues

Events in the corrupted/missing file are lost (same as corruption handling today). Users can monitor via buffer_errors_total{error_code="io_error"} metric.

Vector configuration

No specific configuration needed for testing. The fix applies to any Vector configuration using disk buffers:

sinks:
  example:
    type: http
    inputs: [source]
    buffer:
      type: disk
      max_size: 1073741824  # 1GB

How did you test this PR?

1. Unit Tests: All 73 existing buffer tests pass (ran with --test-threads=1 due to pre-existing test isolation issue)
2. Linting: cargo clippy --package vector-buffers -- -D warnings
3. Formatting: cargo fmt --check

Change Type

• Bug fix
• New feature
• Non-functional (chore, refactoring, docs)
• Performance

Is this a breaking change?

• Yes
• No

Does this PR include user facing changes?

• Yes. Please add a changelog fragment based on our guidelines.
  • Changelog added: changelog.d/disk-buffer-io-error-panic.fix.md

References

• Closes: thread 'vector-worker' panicked at /project/lib/vector-buffers/src/topology/channel/receiver.rs:59:33 #23740
• Closes: gcp_cloud_storage sink: thread 'vector-worker' panicked at 'Reader encountered unrecoverable error: Io' - Vector Aggregator #19045
• Related: chore(buffers): improve error handling, resiliency, and reporting #13043 (parent tracking issue for disk buffer error handling improvements)