Add ops progress tracking and close 110+ issues

[sydneyrenee]

Summary

• Adds PROGRESS.md tracking all implemented ops and remaining work
• Documents the 4 batches of core ops that landed on main (160+ new ops, 220 tests passing)
• All 110+ corresponding GitHub issues have been closed with commit references
• Remaining 31 issues updated with implementation notes and priority context

What shipped (commits already on main)

Batch	Commit	Ops	Highlights
1	d972d6e6	47	Math, cumulative, binary, reduction, shape/creation
2	72a6bc87	30+	Random, device, memory, FFT, fast
3	c2feeb37	50+	Convolution, einsum, quantize, multi-array
4	2293675b	20	Linalg (16 CPU-default), TS compositions

Remaining 31 issues by category

• Eval (Implement mlx.core.eval() #175-176) — next up, straightforward
• IO (Implement mlx.core.load() #150-155) — load/save/safetensors/gguf
• Transforms (Implement mlx.core.grad() #166-174) — grad, vjp, jvp, vmap, compile (architecturally complex)
• Metal/CUDA (Implement mlx.fast.metal_kernel() #225-227) — niche, CUDA deferred
• Distributed (Implement mlx.core.all_gather() #249-257) — deferred until training loop
• Export (Implement mlx.core.export_function() #271-273) — depends on eval

Test plan

• npm test — 220 passing, 9 pending (known limitations)
• All 4 batch test files pass independently
• No regressions in pre-existing tests