[HAL][SD] Fix SD card initialization with 4-bit bus width

[Avlaak]

Problem
When hsd->Init.BusWide is set to SDIO_BUS_WIDE_4B, the SD card fails to initialize. Users are forced to use a workaround: set BusWide = SDIO_BUS_WIDE_1B and manually call HAL_SD_ConfigWideBusOperation() afterward.

This issue was discussed in the ST Community forum:
Running SDIO 4BIT wide mode

The recommended workaround requires users to modify MX_SDIO_SD_Init() and is inconsistent with the STM32H7 HAL behavior where 4-bit mode works out of the box.

Root Cause
In the internal function SD_InitCard(), the SDIO peripheral was reconfigured with user parameters including BusWide:

/* Configure SDIO peripheral interface */
(void)SDIO_Init(hsd->Instance, hsd->Init);

At this point, the SD card has not received the ACMD6 command to switch bus width. This causes desynchronization:

• SDIO controller operates in 4-bit mode
• SD card still operates in 1-bit mode
• All subsequent data transfers fail

Solution
Two changes were made to align STM32F4 HAL behavior with STM32H7 HAL:

1. Fixed SDIO_Init() call in SD_InitCard() to force 1-bit bus width. The SDIO peripheral reconfiguration now explicitly sets 1-bit bus width while preserving other user settings (ClockDiv, ClockEdge, etc.): This ensures the SDIO controller and SD card remain synchronized in 1-bit mode until proper bus width switching via ACMD6.

2. Added automatic bus width configuration in HAL_SD_Init(). After successful card initialization, HAL_SD_Init() now automatically calls HAL_SD_ConfigWideBusOperation() to apply the user-specified bus width, followed by a readiness check:

/* Configure the bus wide */
if (HAL_SD_ConfigWideBusOperation(hsd, hsd->Init.BusWide) != HAL_OK)
{
  return HAL_ERROR;
}

/* Verify that SD card is ready to use after Initialization */
tickstart = HAL_GetTick();
while ((HAL_SD_GetCardState(hsd) != HAL_SD_CARD_TRANSFER))
{
  if ((HAL_GetTick() - tickstart) >=  SDMMC_SWDATATIMEOUT)
  {
    hsd->ErrorCode = HAL_SD_ERROR_TIMEOUT;
    hsd->State = HAL_SD_STATE_READY;
    return HAL_TIMEOUT;
  }
}

This is consistent with the STM32H7 HAL implementation where this exact code exists and works correctly.

Benefits

• No workarounds needed: Users can set BusWide = SDIO_BUS_WIDE_4B in CubeMX and it just works
• Simplified BSP code: BSP_SD_Init() no longer needs to call HAL_SD_ConfigWideBusOperation() manually
• Consistency: STM32F4 HAL now behaves identically to STM32H7 HAL
• Backward compatible: Applications already calling HAL_SD_ConfigWideBusOperation() manually will continue to work (the call is idempotent)