Test flashing speed improvements

[SergioGasquez]

• Update all flasher stubs to https://github.com/espressif/esptool/tree/master/esptool/targets/stub_flasher/2
  • The newer stubs improve buffering, support asynchronous flash operations and blank-sector handling, and boost the flash clock on supported chips.
• Switch to the requested baud rate before uploading the larger stub
• Fall back to an uncompressed flash transfer when deflate output would be larger than the original data
• Add regression coverage for the 32-bit FLASH_END and FLASH_DEFL_END payloads required by current stubs

Benchmarks

Benchmarked on an ESP32-C6 connected through a CP2102N USB-to-UART adapter, flashing the approximately 1.16 MiB
espflash/tests/data/esp_idf_firmware_c6.elf image. Times are wall-clock flash-command durations; lower is better.

Scenario	Existing implementation	This PR	Improvement
Representative run at 921,600 baud	12.609 s	11.516 s	8.7% faster
Correctness-verified run at 921,600 baud	14.650 s	13.460 s	8.1% faster
Pre-erased flash at 921,600 baud	11.317 s	10.495 s	7.3% faster

At 115,200 baud, the new implementation was 0.9% slower because the newer stub is larger and cannot benefit from the early baud-rate change. Most of the improvement is therefore visible at higher baud rates, 921,600 baud is recommended when supported by the adapter and target.

After benchmarking, the flashed contents were checked using target-side MD5 verification and the temporary flash region was cleaned.