i2cdev: Fix i2c param config and driver install order for esp32 target

[AxelLin]

i2c_driver_install() needs to be called before i2c_param_config(). This fixes panic reported on [1].

Link: [1] espressif/esp-idf#10163
Signed-off-by: Axel Lin axel.lin@ingics.com

[UncleRus]

I'm sorry, but I can't merge this PR: it will break compatibility with previous versions of the ESP-IDF. Obviously, we need to either wait for Espressif to respond to issue #10163 (and possibly fix it) or use conditional compilation.

[AxelLin]

I think this won't cause regression because in esp-idf i2c_tools, it calls with below order in all esp-idf versions:
i2c_driver_install()
i2c_param_config() (this is called in i2c_master_driver_initialize()).

e.g.
v4.4: https://github.com/espressif/esp-idf/blob/release/v4.4/examples/peripherals/i2c/i2c_tools/main/cmd_i2ctools.c
master: https://github.com/espressif/esp-idf/blob/master/examples/peripherals/i2c/i2c_tools/main/cmd_i2ctools.c

The problem in current master tree happens if calling i2c_param_config() after i2c_driver_delete().
So probe device function as below will trigger the issue after 1st iteration of the loop.

for_each_i2c_device() {
        i2c_param_config()
        i2c_driver_install()
        ...
        i2c_driver_delete()
}

I believe this is a regression in esp-idf, but I'm not sure when it will be fixed.
I'm fine to wait for espressif's response, just want to clarify this patch should not cause regression.

[UncleRus]

Added check for esp-idf version.
@AxelLin Thanks a lot!