[breaking] Remove the predictor param, allow fallback to prediction using DMatrix. (#9129)

- A `DeviceOrd` struct is implemented to indicate the device. It will eventually replace the `gpu_id` parameter.
- The `predictor` parameter is removed.
- Fallback to `DMatrix` when `inplace_predict` is not available.
- The heuristic for choosing a predictor is only used during training.

jvm-packages/xgboost4j-gpu/src/test/java/ml/dmlc/xgboost4j/gpu/java/BoosterTest.java

@@ -78,7 +78,6 @@ public class BoosterTest {
         put("num_round", round);
         put("num_workers", 1);
         put("tree_method", "gpu_hist");
-        put("predictor", "gpu_predictor");
         put("max_bin", maxBin);
       }
     };

jvm-packages/xgboost4j-spark-gpu/src/main/scala/ml/dmlc/xgboost4j/scala/rapids/spark/GpuPreXGBoost.scala

@@ -281,7 +281,6 @@ object GpuPreXGBoost extends PreXGBoostProvider {
             // - predictor: Force to gpu predictor since native doesn't save predictor.
             val gpuId = if (!isLocal) XGBoost.getGPUAddrFromResources else 0
             booster.setParam("gpu_id", gpuId.toString)
-            booster.setParam("predictor", "gpu_predictor")
             logger.info("GPU transform on device: " + gpuId)
             boosterFlag.isGpuParamsSet = true;
           }