feat(core): enable runaway tool-call rate limiting and daily cost cap by default

[bug-ops]

Summary

• A fresh Zeph install shipped with the tool-call rate limiter disabled and no daily cost cap, leaving no built-in defense against a stuck tool-call loop or unbounded LLM spend. Both enforcement mechanisms (ToolRateLimiter, CostTracker::check_budget()) already existed and were fully wired into the tool-dispatch and LLM-dispatch hot paths — this is a default-value + wizard-prompt + metrics change, not new enforcement logic.
• Fixes a serde split-brain bug found during architecture review: RateLimitConfig.enabled used a bare #[serde(default)] on the field (resolving to false) while impl Default set true, so an absent [security.rate_limit] TOML section was protected but a present-but-partial section (e.g. only one sub-limit configured) silently stayed disabled. Both paths now agree via a named default fn, verified with regression tests at both the leaf RateLimitConfig and nested SecurityConfig level.
• Rate limiter now enabled by default (unchanged per-category thresholds: shell 30/min, web 20/min, memory 60/min, mcp 40/min, other 60/min, 30s cooldown).
• Daily cost cap now defaults to $25.00/day (2500 cents), not unlimited (0).
• Explicit user-set values (enabled = false, max_daily_cents = 0, or any other explicit non-zero value) are preserved unchanged — this only affects sparse/absent config sections.
• Adds a zeph_cost_budget_exhausted_total Prometheus counter mirroring the existing zeph_security_rate_limit_trips metric.
• Adds remediation hints to the budget-exhausted and rate-limit-exceeded user-facing messages (how to raise/disable via config.toml).
• Updates the --init wizard: rate-limit prompt now defaults to enabled; new prompt for the daily cost cap in dollars.
• Adds an optional advisory-only migration step (migrate_rate_limit_advisory) that documents the new defaults for existing configs via commented-out TOML — purely documentary, cannot activate a live value.
• Syncs config/default.toml and book/src/ docs with the new defaults.

Test plan

• cargo +nightly fmt --check
• cargo clippy --profile ci --workspace --all-targets --features "desktop,ide,server,chat,pdf,scheduler,testing" -- -D warnings
• cargo nextest run --config-file .github/nextest.toml --workspace --features "desktop,ide,server,chat,pdf,scheduler" --lib --bins — 14791/14791 passed
• RUSTFLAGS="-D warnings" RUSTDOCFLAGS="--deny rustdoc::broken_intra_doc_links" cargo doc --no-deps --workspace --features "desktop,ide,server,chat,pdf,scheduler"
• New regression tests: absent/partial [security.rate_limit] section resolves enabled = true (the split-brain guard), explicit enabled = false preserved, explicit max_daily_cents = 0/non-zero preserved, CostConfig absent-section default, wizard dollars-to-cents parsing (extracted to a pure fn with unit tests)
• gitleaks protect --staged
• Live-testing playbook (.local/testing/playbooks/security.md, Scenarios 1-6) — not yet run; tracked in coverage-status.md as Untested, to be exercised in a future CI/live-testing cycle

Closes #6469