incin 0.1.0

Machine learning framework for setting fire to dimensionality bugs
Documentation
# {{PROJECT_NAME}}

Scaffolded by `cargo incin new mnist` -- a minimal Incin training loop with
telemetry pre-wired.

## Quickstart

```bash
cargo run
```

In another terminal, watch loss live (no browser, no Python):

```bash
cargo incin watch --run-id <the run id printed on startup>
```

## Next steps

- `src/main.rs` trains on synthetic random data so it runs immediately. Swap
  in real MNIST by following incin's own
  `crates/incin/examples/mnist_training.rs` (loads
  `incin_data::vision::mnist::MnistDataset`).
- Add more `reporter.scalar(...)` / `reporter.gradient_norm(...)` calls
  (see `incin_telemetry::reporter::Reporter`) to track anything else you
  care about -- they show up in `cargo incin watch` automatically.