Add timed coding mock interviews and evidence-based postmortems

[rosshd]

Parent: #56

Depends on: #45, #46, #47, and the interview-profile, problem-catalog, attempt-lifecycle, adaptive-selector, safe-runner, and solution-evaluator children of #56.

Problem

Teaching mode and interview mode have different goals.

A tutor should scaffold, repair gaps, and schedule review.

An interviewer should clarify requirements, observe reasoning, limit intervention, introduce follow-ups, manage time, and defer most feedback until the exercise ends.

No current issue defines that distinct assessment behavior.

Goal

Add an explicit timed mock-interview mode with realistic interviewer interaction, durable coding evidence, and a structured postmortem.

Session setup

• Select target level, language, duration, and focus.
• Choose an eligible unseen or sufficiently spaced problem.
• Explain the assessment rules, timebox, tool boundary, and how to stop.
• Let the learner opt into accessibility accommodations.
• Create a dedicated attempt marked as mock interview.

Interview phases

• Problem introduction and clarification.
• Learner plan and approach discussion.
• Implementation.
• Testing and debugging.
• Complexity explanation.
• One or more bounded follow-up changes when time permits.
• Learner questions and closure.

Phase timing guides the interviewer but does not force abrupt transitions that invalidate useful evidence.

Interviewer policy

• Ask natural clarification questions without creating graded checks accidentally.
• Do not teach, praise excessively, reveal the pattern, or give unsolicited hints.
• If the learner explicitly asks for help, record the intervention and provide only the configured level of assistance.
• Distinguish silence, thinking time, confusion, navigation, and an answer.
• Keep state and timing claims authoritative.
• Allow the learner to stop the interview and optionally switch into teaching mode through an explicit transition.

Postmortem

• Summarize strengths and the highest-leverage improvement areas.
• Separate algorithm choice, correctness, complexity, testing, debugging, communication, and response to follow-ups.
• Cite attempt evidence rather than generic impressions.
• Identify which feedback is certain, uncertain, or based on evaluator judgment.
• Recommend a small next practice sequence through the adaptive selector.
• Do not assign mastery solely from the mock-interview score.

Acceptance criteria

• Mock-interview mode is explicit and cannot start accidentally from normal tutoring.
• Timing, phase, and intervention events are durable and resumable where appropriate.
• The interviewer does not leak hints or teaching content before an explicit help request.
• Clarification dialogue does not pollute graded-answer state.
• Follow-ups modify constraints or requirements without silently replacing the original problem.
• Stopping or switching to teaching mode is explicit and preserves evidence.
• The postmortem references concrete attempt evidence and separates evaluation dimensions.
• Runner or provider failure invalidates affected assessment evidence instead of penalizing the learner.
• Tests cover strong, struggling, silent, help-requesting, interrupted, and tool-failure sessions.
• Live dogfood includes at least two target levels and records human review of realism.
• make check passes.

Out of scope

• Human interviewer marketplaces.
• Live video or audio analysis in the first version.
• System-design and behavioral mock interviews.
• Claims that one mock score predicts hiring outcomes.

[rosshd]

The manual placement review confirmed that full interview realism should remain an explicit mode instead of making onboarding carry every mock-interview phase.

• Build an explicit timed mock-interview session controller #73 defines the implementable timed session-controller slice for this issue.
• It depends on the compact conversation (Replace placement questionnaires with a compact interviewer conversation #69), verified interview editor (Add a verified Neovim interview mode without autocomplete #70), and evidence-driven debrief/result work (Add one evidence-driven debrief and a course-start passport #71).
• Quick placement targets a 10-12 minute routing sample. Full mock retains clarification, plan communication, constrained coding, complexity discussion, follow-up adaptation, and an evidence-based postmortem.