Add EdTech scenario: student-teacher memory tracking

[Neal006]

Context

MemoryLens currently benchmarks generic personal facts (name, city, age). This issue adds a domain-specific scenario targeting EdTech — the most natural real-world use case for multi-turn AI memory.

What to build

Create `simulator/edtech_facts.py` with a new fact set:

```python
EDTECH_FACTS = [
Fact("student_name", "Priya Sharma", injected_at=0),
Fact("grade", "10th standard", injected_at=1),
Fact("weak_subject", "quadratic equations", injected_at=2,
updated_at=50, updated_value="trigonometry"), # student improves
Fact("learning_style", "visual", injected_at=3),
Fact("last_test_score", "68 percent", injected_at=5,
updated_at=40, updated_value="79 percent"), # score improves
Fact("preferred_language", "Hindi", injected_at=6),
Fact("upcoming_exam", "CBSE Board", injected_at=8),
Fact("daily_study_hours", "2 hours", injected_at=10,
updated_at=70, updated_value="3 hours"),
]
```

Also add EdTech-flavoured filler turns in `simulator/conversation.py`:

```python
EDTECH_FILLER = [
"Can you explain the Pythagorean theorem?",
"What is the formula for compound interest?",
"I don't understand photosynthesis.",
"Can you give me practice problems for algebra?",
...
]
```

Wire it into the CLI

```bash
python main.py --scenario edtech --backends naive rag cascading
```

Why this matters

VedaAI, Byju's, and similar EdTech companies use AI tutors with multi-session memory. A student's weak topics, scores, and learning style change over the school year — this is exactly the temporal drift problem MemoryLens measures.

Acceptance criteria

• `EDTECH_FACTS` defined in `simulator/edtech_facts.py`
• `--scenario edtech` flag works in `main.py`
• Dashboard has a scenario selector dropdown
• At least 3 filler question categories (math, science, language)
• One test in `tests/test_pipeline.py` using the EdTech scenario

Estimated effort

2 hours. Mostly data definition — no new algorithms.