LICM pass (#87)

* LICM pass

* Uses loop analysis to detect loop tree
* For each loop (starting with the inner ones), create a pre-header and move there loop-invariant instructions
* An instruction is loop invariant if it does not use as argument a value defined earlier in the loop
* File tests to check LICM's correctness
* Optimized pre-header creation
If the loop already has a natural pre-header, we use it instead of creating a new one.
The natural pre-header of a loop is the only predecessor of the header it doesn't dominate.

filetests/licm/multiple-blocks.cton

@@ -0,0 +1,46 @@
+test licm
+
+function multiple_blocks(i32) -> i32 {
+
+ebb0(v0: i32):
+    jump ebb1(v0)
+
+ebb1(v10: i32):
+    v11 = iconst.i32 1
+    v12 = iconst.i32 2
+    v13 = iadd v11, v12
+    brz v10, ebb2(v10)
+    v15 = isub v10, v11
+    brz v15, ebb3(v15)
+    v14 = isub v10, v11
+    jump ebb1(v14)
+
+ebb2(v20: i32):
+    return v20
+
+ebb3(v30: i32):
+    v31 = iadd v11, v13
+    jump ebb1(v30)
+
+}
+; sameln:function multiple_blocks(i32) -> i32 {
+; nextln: ebb0(v0: i32):
+; nextln:     v2 = iconst.i32 1
+; nextln:     v3 = iconst.i32 2
+; nextln:     v4 = iadd v2, v3
+; nextln:     v9 = iadd v2, v4
+; nextln:     jump ebb1(v0)
+; nextln: 
+; nextln: ebb1(v1: i32):
+; nextln:     brz v1, ebb2(v1)
+; nextln:     v5 = isub v1, v2
+; nextln:     brz v5, ebb3(v5)
+; nextln:     v6 = isub v1, v2
+; nextln:     jump ebb1(v6)
+; nextln: 
+; nextln: ebb2(v7: i32):
+; nextln:     return v7
+; nextln: 
+; nextln: ebb3(v8: i32):
+; nextln:     jump ebb1(v8)
+; nextln: }