Add bulk file_read and bulk memory_read abilities

[bradvin]

Problem

Multi-file and multi-memory retrieval currently requires many sequential tool calls, increasing round count and latency.

Dependencies

• Depends on memory read abilities from Add abilities to read long-term and short-term memory files #28.

Implementation Plan

1) Bulk file read ability

Add file_read_bulk ability:

• Input: paths[] (bounded array, e.g. max 25), optional max_chars_per_file.
• Behavior: attempts each path independently using existing file-read resolution rules (agent file first, workspace fallback).

2) Bulk memory read ability

Add memory_read_bulk ability:

• Input: array of targets, e.g.
  • type=long_term
  • type=short_term + optional date.
• Optional max_chars_per_item.

3) Result envelope

Return deterministic per-item records:

• input (requested key/path)
• success
• error (code/message when failed)
• data (on success)

4) Limits and safety

• Enforce request size and item-count limits.
• Truncate large payloads per item and include truncation metadata.
• Keep ability readonly/idempotent.

5) Reuse existing helpers

• File operations: reuse existing file-read logic/helper paths.
• Memory operations: reuse Memory_Helper read methods.

Acceptance Criteria

• A single tool call can fetch multiple files/memory targets.
• Partial failures do not fail the entire call.
• Output ordering matches input ordering.
• Bounded limits prevent oversized responses.

Test Plan

• Unit tests:
  • all-success batch,
  • mixed success/failure batch,
  • limit enforcement,
  • deterministic ordering.
• Tool declaration and mapping tests.

Out of Scope

• Cross-item dedup or caching.
• Streaming large binary payloads.