Add interactive model experiment suites and prepare 0.1.1

[jpn--]

Changes

Model repositories can now provide self-contained benchmark suites that download checksum-verified data through ActivitySim's shared cache, resolve a current GitHub branch or PR head to an exact source commit, and prompt for typed experiment inputs.

abench /path/to/model discovers YAML suites in .abench, asks which experiment to run, then collects its inputs. Defaults can be accepted with Enter; required or invalid values prompt again. Explicit --set values and --non-interactive support automation. Suite metadata records the original instructions, effective inputs, and resolved commits.

Long builds and model phases now print periodic elapsed-time updates, with current and peak cgroup memory when available. Full logs remain on disk.

Prepares release 0.1.1 and updates the documentation and SANDAG example.

Validation

• 146 unit tests passed locally.
• Pre-commit lint and formatting checks passed.
• Four Docker integration tests passed during implementation, covering serial/MP runs, cache retries, shared data preparation, and typed overrides.
• Interactive input and directory selection verified in a real terminal.
• GitHub CI checks Python 3.10–3.12, built wheels, and Docker integration before release.