remove overzealous check during HEAD() (#19940)

due to a historic bug in CopyObject() where
an inlined object loses its metadata, the
check causes an incorrect fallback verifying
data-dir.

CopyObject() bug was fixed in ffa91f9794 however
the occurrence of this problem is historic, so
the aforementioned check is stretching too much.

Bonus: simplify fileInfoRaw() to read xl.json as well,
also recreate buckets properly.

cmd/global-heal.go

@@ -163,7 +163,10 @@ func (er *erasureObjects) healErasureSet(ctx context.Context, buckets []string,
 	}
 
 	for _, bucket := range healBuckets {
-		_, err := objAPI.HealBucket(ctx, bucket, madmin.HealOpts{ScanMode: scanMode})
+		_, err := objAPI.HealBucket(ctx, bucket, madmin.HealOpts{
+			Recreate: true,
+			ScanMode: scanMode,
+		})
 		if err != nil {
 			// Log bucket healing error if any, we shall retry again.
 			healingLogIf(ctx, err)
@@ -262,6 +265,7 @@ func (er *erasureObjects) healErasureSet(ctx context.Context, buckets []string,
 		// Heal current bucket again in case if it is failed
 		// in the beginning of erasure set healing
 		if _, err := objAPI.HealBucket(ctx, bucket, madmin.HealOpts{
+			Recreate: true,
 			ScanMode: scanMode,
 		}); err != nil {
 			// Set this such that when we return this function
@@ -513,7 +517,7 @@ func (er *erasureObjects) healErasureSet(ctx context.Context, buckets []string,
 			// we let the caller retry this disk again for the
 			// buckets it failed to list.
 			retErr = err
-			healingLogIf(ctx, err)
+			healingLogIf(ctx, fmt.Errorf("listing failed with: %v on bucket: %v", err, bucket))
 			continue
 		}