fix: pin num-traits to 0.2.19 to match platform-num's exact version

[konard]

Summary

Fixes #26

The issue asks whether the direct num-traits dependency in platform-trees can be removed since platform-num already depends on it internally.

Root Cause Analysis

Direct removal is not possible. Rust requires explicit direct dependencies — transitive deps cannot be used in code. Attempting to remove num-traits from Cargo.toml yields:

error[E0432]: unresolved import `num_traits`
 --> src/link_type.rs:1:5
  |
1 | use num_traits::{FromPrimitive, Unsigned};
  |     ^^^^^^^^^^ use of unresolved module or unlinked crate `num_traits`

The real risk: version drift. Before this fix:

• platform-trees: num-traits = "0.2" (any 0.2.x)
• platform-num v0.5.0: num-traits = "0.2.19" (pinned)

If these ever resolved to different num-traits versions, Unsigned and FromPrimitive would be incompatible types from the caller's perspective and compilation would fail.

Changes

1. Cargo.toml: Pin num-traits from "0.2" → "0.2.19" to match platform-num's exact version, ensuring version consistency.

2. docs/case-studies/issue-26/README.md: Full case study with timeline, root cause analysis, proposed solutions, and references.

3. changelog.d/...: Patch-level changelog fragment.

Future Path to Full Removal

Full removal of the num-traits dependency would require platform-num to re-export the traits it uses internally (Unsigned, FromPrimitive, etc.). I filed linksplatform/Numbers#141 requesting this change. Once that's implemented and a new version of platform-num is released, platform-trees could drop num-traits entirely.

Testing

All 32 existing tests pass with the pinned version:

test result: ok. 32 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out

---

This PR was created automatically by the AI issue solver

[konard]

🤖 Solution Draft Log

This log file contains the complete execution trace of the AI solution draft process.

💰 Cost: $0.830036

📊 Context and tokens usage:

Claude Sonnet 4.6:

• Context window: 50.9K / 1M (5%) input tokens, 11.3K / 64K (18%) output tokens

Total: (40.3K + 1.4M cached) input tokens, 11.3K output tokens, $0.745098 cost

Claude Haiku 4.5:

Total: (32.3K + 250.3K cached) input tokens, 3.9K / 64K (6%) output tokens, $0.084938 cost

🤖 Models used:

• Tool: Anthropic Claude Code
• Requested: sonnet
• Main model: Claude Sonnet 4.6 (claude-sonnet-4-6)
• Additional models:
  • Claude Haiku 4.5 (claude-haiku-4-5-20251001)

📎 Log file uploaded as Gist (825KB)

• View complete solution draft log

---

Now working session is ended, feel free to review and add any feedback on the solution draft.