Hash a model's bytes once, at open, not on every acquire

A border fill acquires the filler once a tile, and each acquire hashed
the 28 MB model twice — 60 ms a tile, a third of the tile's run on a
throttled TensorRT. The Model keeps its hash from open.

core/dr-inference-engine/src/engines.rs

@@ -29,7 +29,12 @@ pub fn hash(bytes: &[u8]) -> u64 {
 
 /// The cache entry for `bytes` compiled on `rung`.
 pub fn key(rung: Rung, bytes: &[u8]) -> String {
-    format!("{}:{:016x}", rung.label(), hash(bytes))
+    key_of(rung, hash(bytes))
+}
+
+/// The same, from a hash already taken.
+pub fn key_of(rung: Rung, hash: u64) -> String {
+    format!("{}:{:016x}", rung.label(), hash)
 }
 
 /// Where QNN's compiled context for `bytes` lives.