Add documentation for ssh, describe, completion commands,
--output flag, structured output examples, and --wait flag.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Add three new command groups:

- `verda images` — browse OS images with --type and --category
  filters. Shows name, category, and details (e.g., CUDA version).

- `verda availability` — check instance type availability per
  location. Supports --location, --type, and --spot flags.
  Single-type check returns a boolean; full matrix shows all
  available types per datacenter.

- `verda cost` — cost estimation, pricing, and billing:
  - `estimate` — calculate hourly/daily/monthly costs for an
    instance type + optional volumes, with spot pricing support
  - `price-history` — show historical fixed and dynamic pricing
  - `balance` — display current account balance

All commands support --output json/yaml for scripting.
Includes tests for cost math, formatting, and filtering logic.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

The /instance-types/{type} endpoint returns 404. Fetch all instance
types and filter client-side instead.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

The Verda API requires lowercase 'usd', not 'USD'.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

The /instance-types/price-history endpoint has been removed by the
API (410 Gone). Dynamic pricing is no longer supported.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

List available datacenter locations with code, name, and country.
Supports --output json/yaml.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

- `verda cost running` shows burn rate of all running instances
  with per-instance breakdown including attached volume costs
- Add scripts/live-test.sh for automated integration testing

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

When --output is json or yaml, return nil from Factory.Status()
so spinners don't write escape codes to stdout and break piping.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

- `verda instance-types` lists all types with specs and pricing,
  with --gpu, --cpu, and --spot filters. Groups output by GPU vs CPU.

- `verda vm list` now prints a static table when stdout is piped
  or redirected, instead of launching the interactive picker.
  Interactive mode still works in a terminal.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

- Strip duplicate count prefix from API GPU descriptions
  (e.g., "1x 1x H100 SXM5 80GB" → "1x H100 SXM5")
- Strip trailing VRAM from description (shown in separate column)
- Format large RAM values as TB (e.g., 1440GB → 1.4TB)
- Widen GPU column for longer names

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Compute max hostname length so columns stay aligned with long names.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

- Separate "Flags" from "Global Flags" in help output
- Register --output flag completion values (json/yaml/table)
- Remove dead JSON bool field from vm list options
- Fix Poll() timeout to return error instead of nil
- Consolidate duplicate polling code from status_view.go into shared wait.go
- Centralize status messages and terminal statuses in status_messages.go
- Add interactive picker for verda ssh (no-arg shows running instances)
- Add interactive picker for vm describe and volume describe (no-arg shows list)
- Add reusable TestFactory in cmd/util/testing.go for command-level tests
- Add arg-validation and status message tests
- Fix pre-existing lint issues (stuttering type names, hugeParam, goimports)
- Ignore .ai/notes/ and docs/plans/ in .gitignore

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

- Version: show human-readable text by default, JSON only with -o json
- Image: show image slug instead of volume name, append OS name if different
- Compute: strip duplicate "1x" prefix from GPU description
- Pricing: remove redundant field (only fixed pricing exists)
- Storage Status: show actual status (attached/detached) with "(OS)" label

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Full build details still available via verda version -o json.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Joins instance type specs, pricing, and live availability data into a
single view — sorted by price, filterable by location, type, kind, and
spot. Addresses user request for a native alternative to the Python
availability script.

Usage:
  verda vm availability
  verda vm availability --location FIN-01 --kind gpu --spot
  verda vm availability -o json

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

- ssh.go: suppress G204 for intentional syscall.Exec into ssh
- images_test.go, instancetypes_test.go: use empty slice instead of nil
  to avoid G602 false positive on nil range

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>