Usb error more informative to user. Refactored vartypes of device's serial.

[theg4sh]

A little change to inform user about problems during device search or open.

[theg4sh]

Please, review the last changes.

[xor-gate]

I'm closing this PR as it is not 100% what we want because there is a mixture of hex/binary serial numbers and needs to be resolved in a generic way by writing a hex2bin/bin2hex implementation or using a existing one (e.g https://raw.githubusercontent.com/xor-gate/stlink2/master/src/utils/hexstr.c).

[theg4sh]

@xor-gate, got it.
I did the same conclusion earlier, so, that's why I didn't push any commits.

Could you, please, take a look for branch https://github.com/theg4sh/stlink/tree/libusb_minimize
There I did changes with serial, so, it structured and always converted to binary and then compared.
Functions: stlink_serial_convert, stlink_serial_init, stlink_serial_clear.

Also, I found that all usb devices walked by 3 times per tool run, e.g. 'st-info --probe', so, did change regarding this - in branch stlink devices initialized once and then used prepared pointer to get next stlink device.
Functions used to iterate stlink devices: stlink_init_devices, stlink_device_next, stlink_devices_rewind and stlink_free_devices.
So, it could be possible to clear sources from non-DRY code.
Please, note, that branch is not clear now.

In other hand, devices iterated for both versions - v1 and v2. I guess, it could be good for DRY, but stlinkv1 uses similar, but not the same backend callbacks.
I want to know your opinion regarding that modifications and will continue to work on it if you approved it.

[xor-gate]

Looks good why so many info is necessary if you normalize to just a string (ascii, or asciihex). Directly converting inline after probe would be more friendly from API point of view.

+    enum stlink_serial_format_type {
+        STSERIALF_UNKNOWN=0,
+        STSERIALF_BINARY,
+        STSERIALF_HEX,
+        STSERIALF_ASCII
+    };