docs: Update KEP with DynamicAllocation API changes

Author: Shekharrajak

docs/proposals/107-spark-client/README.md

@@ -145,7 +145,7 @@ if status.state != ApplicationState.COMPLETED:
 **So that** I can efficiently prepare data for model training.
 
 ```python
-from kubeflow.spark import BatchSparkClient, OperatorBackendConfig
+from kubeflow.spark import BatchSparkClient, OperatorBackendConfig, DynamicAllocation
 
 config = OperatorBackendConfig(
     namespace="ml-jobs",
@@ -158,9 +158,10 @@ response = client.submit_application(
     main_application_file="s3a://ml/features/extract.py",
 
     # Dynamic allocation for cost optimization
-    enable_dynamic_allocation=True,
-    min_executors=5,
-    max_executors=50,
+    dynamic_allocation=DynamicAllocation(
+        min_executors=5,
+        max_executors=50,
+    ),
 
     # Resource configuration
     driver_cores=4,
@@ -345,7 +346,7 @@ response = client.submit_application(
 #### Advanced Features: Dynamic Allocation, Volumes, GPU
 
 ```python
-from kubeflow.spark import BatchSparkClient, OperatorBackendConfig
+from kubeflow.spark import BatchSparkClient, OperatorBackendConfig, DynamicAllocation
 
 config = OperatorBackendConfig(namespace="default")
 client = BatchSparkClient(backend_config=config)
@@ -360,10 +361,11 @@ response = client.submit_application(
     executor_memory="8g",
 
     # Enable dynamic allocation (auto-scaling)
-    enable_dynamic_allocation=True,
-    initial_executors=2,
-    min_executors=1,
-    max_executors=10,
+    dynamic_allocation=DynamicAllocation(
+        initial_executors=2,
+        min_executors=1,
+        max_executors=10,
+    ),
 
     # Configure volumes for data access
     volumes=[{
@@ -771,6 +773,41 @@ class ApplicationStatus:
     message: Optional[str]
 ```
 
+### API Changes: Dynamic Allocation
+
+**Breaking Change in v0.1.0**: Dynamic allocation now uses a configuration object instead of scattered parameters.
+
+**Old Pattern (Deprecated):**
+```python
+client.submit_application(
+    enable_dynamic_allocation=True,
+    initial_executors=2,
+    min_executors=1,
+    max_executors=10,
+)
+```
+
+**New Pattern:**
+```python
+from kubeflow.spark import DynamicAllocation
+
+client.submit_application(
+    dynamic_allocation=DynamicAllocation(
+        initial_executors=2,
+        min_executors=1,
+        max_executors=10,
+    ),
+)
+```
+
+**Key Changes:**
+- Import `DynamicAllocation` from `kubeflow.spark`
+- Pass as single `dynamic_allocation` parameter
+- `None` or omit parameter = disabled
+- Object presence = enabled (no `enabled` field)
+- Validation: `min_executors ≤ initial_executors ≤ max_executors`
+- Default values: `initial_executors` from `num_executors`, `min_executors=1`, `max_executors=num_executors*2`
+
 ### Impact Analysis
 
 #### Impact on Data Engineers