Implement SwidenLow and SwidenHigh for the interpreter

Implemented `SwidenLow` and `SwidenHigh` for the Cranelift interpreter,
doubling the width and halving the number of lanes preserving the low
and high halves respectively.

Conversions are performed using signed extension.

Copyright (c) 2021, Arm Limited

cranelift/filetests/filetests/runtests/simd-swidenhigh.clif

@@ -0,0 +1,26 @@
+test interpret
+test run
+target aarch64
+set enable_simd
+target x86_64
+
+function %swidenhigh_i8x16(i8x16) -> i16x8 {
+block0(v0: i8x16):
+    v1 = swiden_high v0
+    return v1
+}
+; run: %swidenhigh_i8x16([1 -2 3 -4 5 -6 7 -8 9 -10 11 -12 13 -14 15 -16]) == [9 -10 11 -12 13 -14 15 -16]
+
+function %swidenhigh_i16x8(i16x8) -> i32x4 {
+block0(v0: i16x8):
+    v1 = swiden_high v0
+    return v1
+}
+; run: %swidenhigh_i16x8([1 -2 3 -4 5 -6 7 -8]) == [5 -6 7 -8]
+
+function %swidenhigh_i32x4(i32x4) -> i64x2 {
+block0(v0: i32x4):
+    v1 = swiden_high v0
+    return v1
+}
+; run: %swidenhigh_i32x4([1 -2 3 -4]) == [3 -4]

cranelift/filetests/filetests/runtests/simd-swidenlow.clif

@@ -0,0 +1,26 @@
+test interpret
+test run
+target aarch64
+set enable_simd
+target x86_64
+
+function %swidenlow_i8x16(i8x16) -> i16x8 {
+block0(v0: i8x16):
+    v1 = swiden_low v0
+    return v1
+}
+; run: %swidenlow_i8x16([1 -2 3 -4 5 -6 7 -8 9 -10 11 -12 13 -14 15 -16]) == [1 -2 3 -4 5 -6 7 -8]
+
+function %swidenlow_i16x8(i16x8) -> i32x4 {
+block0(v0: i16x8):
+    v1 = swiden_low v0
+    return v1
+}
+; run: %swidenlow_i16x8([1 -2 3 -4 5 -6 7 -8]) == [1 -2 3 -4]
+
+function %swidenlow_i32x4(i32x4) -> i64x2 {
+block0(v0: i32x4):
+    v1 = swiden_low v0
+    return v1
+}
+; run: %swidenlow_i32x4([1 -2 3 -4]) == [1 -2]