rework fsdev driver, fix ch32v203 race condition and stability issue

[hathach]

Describe the PR
More effor and finally able to fix ch32v203 race condition especially with windows host. ch32v203 behaves a bit different comparing to stm32 usb

• correctly handle setup prepare at dcd_edpt0_status_complete(), which fixes the race condition with windows where we could miss setup packet (setup bit set, but count = 0)
• improve dcd int handler
  • skip DIR and use CTR TX/RX to handle complete transfer
  • clear CTR first, except for setup which we need to get data first
  • separate handle_ctr_setup()
• fix v203 race condition between rx bufsize and RX_STAT which cause PMAOVR, occurs with WRITE10 (msc command out then write10 data out).
• fix ch32v203 seems to unconditionally accept ZLP on EP0 OUT, which can incorrectly use queued_len of previous transfer. So reset total_len and queued_len to 0.
• fsdev improve ep bit manipulation
• read/write packet enhancement, merge 16-bit and 32-bit together
• merge read/write packet fifo (16-bit and 32-bit)
• also add iar kickstart stm32f103ze board
• hil: add nanoch32v203 to hil pool, also run hil tests in parallel

[hathach]

new release pico-sdk break the ci, merge and fixedd that later in separated PR.