docs: reframe getting-started as Use as a library; add Vision section

[ammar-s847]

Summary

Two related README changes:

1. Reframe "Getting started" → "Use as a library". This repo is the library, so the framing matches. Folds the separate Rust / Python subsections into one section: install snippet, Python example first (most likely entry point), Rust example below. The on-disk disk feature is split out as its own snippet.

2. Add a Vision section that situates hebb in the existing computational-neuroscience and SNN ecosystem and lists concrete interop directions. Briefly:

• PyTorch-based SNN ML frameworks: snnTorch, BindsNET, Norse, SpikingJelly, Rockpool (surrogate-gradient training).
• Biophysical simulators: NEST, Brian2, NEURON, GeNN, Arbor.
• Neuromorphic platforms: Intel Loihi / Lava, SpiNNaker, BrainChip Akida, SynSense.

Concrete roadmap items called out:

• PyTorch interop (train in snnTorch / Norse / SpikingJelly, export to hebb)
• NeuroML / SONATA / PyNN model-description importers
• NWB experimental-data ingestion
• Event-camera datasets (DVS / N-MNIST / N-Caltech101 / DVS-Gesture)
• Neuromorphic targets (Loihi via Lava IR, SpiNNaker, Akida)
• More neuron + plasticity models (multi-compartment, e-prop, equilibrium propagation, calcium-based plasticity)
• Performance (SIMD, wgpu GPU backend, deterministic parallel ticking)
• WASM (browser simulation)

Also adds a row in "Who is this for?" for PyTorch / SNN-ML researchers, pointing them at the Vision section.

Test plan

• Render README preview on GitHub
• Confirm all section anchor links resolve