GPU backend

Adds GPU backend and "CUDA" feature to all the tests. As of MR creation, there were no forks in tested IRs so only vanilla non-parallel (single block, single thread) codegen has been tested.

juno_samples/test2.jn

-#[entry]
-fn main<m, n : usize>() -> i32[m, n, 64, 64] {
-    let res : i32[m, n, 64, 64];
-    for bi = 0 to m {
-        for bj = 0 to n {
-            let tile : i32[64, 64];
-            for ti = 0 to 64 {
-                for tj = 0 to 64 {
-                    tile[ti, tj] = (ti as i32) + (tj as i32) + (bi as i32) + (bj as i32);
-                }
-            }
-            for si = 1 to 63 {
-                for sj = 1 to 63 {
-                    tile[si, sj] = tile[si-1, sj-1] + tile[si+1, sj+1];
-                }
-            }
-            for ri = 0 to 64 {
-                for rj = 0 to 64 {
-                    res[bi, bj, ri, rj] = tile[ri, rj];
-                }
-            }
-        }
-    }
-    return res;
-}