feat: add analyze-graph for execution-context recommendations

[dapi]

Problem

After an Execution Handoff is available, users need a compact way to inspect the explicit execution-context graph and identify useful context recommendations.

Outcome

Add a read-only command:

memory-bank-cli analyze-graph \
  --handoff .memory-bank/handoffs/FT-042.json

The command analyses typed, explicit evidence from handoffs and reports:

• a recommended context set for continuing the task;
• high-fan-out documents and artefacts for review;
• repeatedly co-accessed documents, decisions, files, and verification artefacts;
• unresolved or weakly evidenced links.

Constraints

• Analyse typed evidence only; do not collapse derived_from, navigation, Git, test, and session relations into one unqualified weight.
• Preserve a source reference and relation type for every finding.
• Do not automatically edit, move, split, or merge Memory Bank documents.
• No graph database or LLM inference in the initial implementation.
• Recommendations are advisory and must be explainable.

Acceptance

• The command accepts an Execution Handoff JSON artefact produced by handoff build.
• Output distinguishes findings, recommendations, and evidence.
• Each recommendation lists its contributing nodes, relation types, and sources.
• Fixtures cover strong co-access, high fan-out, and unresolved-link cases.
• Markdown and JSON output are available.

Dependencies

• docs: define evidence-backed execution handoff contract memory-bank#95 defines the handoff contract.
• feat: build a read-only execution handoff from explicit project evidence #47 supplies the handoff builder.
• research: evaluate observed co-access signals for context recommendations memory-bank#96 evaluates which observed signals are reliable enough for recommendations.

[dapi]

Pull request: #50

Verification performed:

• go test -count=1 -race ./...
• go vet ./...
• git diff --check
• successful CLI build and JSON smoke test for the high-fan-out fixture