kernel_patch_verify: Introduce AI_LOG_ALWAYS variable handling

Introduce AI_LOG_ALWAYS variable that user can configure in
environment to force the "AI thought process" for all patches - this
is useful for reviewing the research that AI has done about the
patch even if the patch itself is fine - typically this includes the
checklists of previous review of the patch and if they have been
addressed etc.

Signed-off-by: Nishanth Menon <nm@ti.com>

Author: nmenon

README.md

@@ -245,6 +245,11 @@ kpv -V -L -A -1
 kpv -V -L -A -b base_branch -t test_branch
 ```
 
+NOTE: You can set the environment variable `AI_LOG_ALWAYS` to 1 to force the
+AI review logic to be available in the log always - even if the patch is fine
+from AI review perspective. Default is to dump the AI thought process only if
+there are review comments to provide for the patch.
+
 ### Semcode Integration
 
 The AI review works best with semcode indexing, which provides better context

kernel_patch_verify

@@ -523,6 +523,14 @@ series_ai_review() {
 					echo "Severity: $severity - $severity_explain"
 					echo "AI-authorship: $ai_authorship"
 					echo "========================================"
+
+				} >> "$LOG_DIR/ai-review.log"
+			else
+				echo "$short_sha: No issues found" >> "$LOG_DIR/ai-review.log"
+			fi
+
+			if [ -f "$worktree_dir/review-inline.txt" -o -n "$AI_LOG_ALWAYS" ]; then
+				{
 					echo ""
 
 					echo "--- Review Comments ---"
@@ -535,8 +543,6 @@ series_ai_review() {
 						echo ""
 					fi
 				} >> "$LOG_DIR/ai-review.log"
-			else
-				echo "$short_sha: No issues found" >> "$LOG_DIR/ai-review.log"
 			fi
 
 			# Clean up the worktree created by review_one.sh