feat: array slicing and concatenation (V16)

[mohammadfawaz]

Adds native array slicing and concatenation, gated to ConsensusVersion::V16. Additive and backwards-compatible.

Slice — r[a..b] range access (Access::Range)

Half-open, constant u32 bounds, yields [T; b-a]. Read-only; composes with further accesses.

// r0 : [u8; 5]
slice:        r0[1u32..4u32]        // -> [u8; 3]
nested:       r0[1u32..3u32]        // [[u8;2];4] -> [[u8;2];2]  (outer dim)
compose:      r0[1u32..4u32][0u32]  // slice, then index -> u8

Concat — flattening cast

When casting to [T; n], an operand of type T is one element; an operand of type [T; k] is flattened into its k elements (one level). Whole-match takes priority, so the target type decides the shape.

// concat: [u8;2] ++ [u8;3] -> [u8;5]
cast r0 r1 into r2 as [u8; 5u32];

// prepend a scalar: u8 ++ [u8;3] -> [u8;4]
cast r0 r1 into r2 as [u8; 4u32];

// whole-match (no flatten): three [u8;2] -> [[u8;2];3]
cast r0 r1 r2 into r3 as [[u8; 2u32]; 3u32];

Composition

// remove element at index 2 from [u8;5]
cast r0[0u32..2u32] r0[3u32..5u32] into r1 as [u8; 4u32];

Tests in synthesizer/src/vm/tests/test_v16/.