drivers: spi: ambiq_spid: fix runtime PM leaks and return status

In spi_ambiq_transceive(), three issues affected the runtime PM path
and error reporting:
1. When spi_config() failed, the function returned directly without
   releasing the runtime PM reference acquired by pm_device_runtime_get(),
   causing the usage count to permanently leak.
2. The return status of spi_ambiq_xfer() was overwritten by the return
   value of pm_device_runtime_put_async(), masking transfer errors from
   callers.
3. When pm_device_runtime_get() failed, execution continued into the
   hardware routine instead of exiting early.

Restructure spi_ambiq_transceive() to return immediately on get
failures, route spi_config() errors through a shared cleanup label, and
use a separate variable for the async put return value.

Fixes #116324

Signed-off-by: harshit kudhial <harshitkudhial@gmail.com>

Author: harshitnub077

drivers/spi/spi_ambiq_spid.c

@@ -276,6 +276,7 @@ static int spi_ambiq_transceive(const struct device *dev, const struct spi_confi
 				const struct spi_buf_set *rx_bufs)
 {
 	struct spi_ambiq_data *data = dev->data;
+	int pm_ret;
 	int ret;
 
 	if (!tx_bufs && !rx_bufs) {
@@ -285,29 +286,30 @@ static int spi_ambiq_transceive(const struct device *dev, const struct spi_confi
 	ret = pm_device_runtime_get(dev);
 	if (ret < 0) {
 		LOG_ERR_PM_DEVICE_RUNTIME_GET(dev, ret);
+		return ret;
 	}
 
 	/* context setup */
 	spi_context_lock(&data->ctx, false, NULL, NULL, config);
 
 	ret = spi_config(dev, config);
 	if (ret) {
-		spi_context_release(&data->ctx, ret);
-		return ret;
+		goto end;
 	}
 
 	spi_context_buffers_setup(&data->ctx, tx_bufs, rx_bufs, 1);
 
 	ret = spi_ambiq_xfer(dev, config);
 
+end:
 	spi_context_release(&data->ctx, ret);
 
 	/* Use async put to avoid useless device suspension/resumption
 	 * when doing consecutive transmission.
 	 */
-	ret = pm_device_runtime_put_async(dev, K_MSEC(2));
-	if (ret < 0) {
-		LOG_ERR_PM_DEVICE_RUNTIME_PUT(dev, ret);
+	pm_ret = pm_device_runtime_put_async(dev, K_MSEC(2));
+	if (pm_ret < 0) {
+		LOG_ERR_PM_DEVICE_RUNTIME_PUT(dev, pm_ret);
 	}
 
 	return ret;