Refactor number[] to Float32Array/Float64Array for performance

[Copilot]

Replace all number[] vector types with typed arrays (Float32Array / Float64Array) for better memory layout, computation efficiency, and typed array API compatibility.

Changes

• FilterDataArray union type — Float32Array | Float64Array, exported for consumer use
• Generic filter<T>() — return type matches input type via generics
• createTypedArray helper — dynamically constructs the correct typed array type from the input, used for all internal buffer allocation
• xPrev / dxPrev — changed from number[] | null to FilterDataArray | null; initialization now copies input via typed array constructor instead of aliasing
• Type declarations — updated both types/OneEuroFilter.d.ts and types/src/OneEuroFilter.d.ts
• README — documents accepted input types
• Example — updated to use new Float32Array([...])

Usage

import { OneEuroFilter } from "@webarkit/oneeurofilter-ts";

const filter = new OneEuroFilter(0.0001, 0.01);

// Float32Array — returns Float32Array
const result32 = filter.filter(Date.now(), new Float32Array([1.0, 2.0, 3.0]));

// Float64Array — returns Float64Array (when higher precision needed)
filter.reset();
const result64 = filter.filter(Date.now(), new Float64Array([1.0, 2.0, 3.0]));