GUI/OptionsDialog: Fix proxy input validation feedback (fixes #166)

[bigshiny90]

Fixes #166

QValidatedLineEdit:

1. Added setAllowEmptyInput() to QValidatedLineEdit to allow it to treat empty values as valid or invalid. Required for empty proxy values to be considered not valid. Default is true to conform to other usage in the code.
2. Added setAllowValidationWhileEditing to allow for live validation while editing (default is false, to not break other usage in code). Previously it only validated on FocusOut. This was required for OptionsDialog proxy settings otherwise edge-case where editing a proxy field and then clicking OK (without explicitly changing focus beforehand) resulted in no validation and acceptance of invalid values.
3. Removed forcing error_locations after calling validate with validator's that do not actually support error_locations. (Bitcoin address validator is explicitly handled and does support error locations)
4. clears invalid styling when disabling

OptionsDialog

1. Fixed QValidatedLineEdit Socks Proxy and Tor proxy input validation to properly validate and show visual feedback. These input fields are QValidatedLineEdit instances but were not previously validating the input (fully) or giving validation feedback. Empty values were previously considered valid but this has been fixed. Added validation of IPv4 style input using inet_pton
2. Added setAllowEmptyInput(false) to proxy fields to allow them to treat empty values as valid or invalid. Required for empty proxy values to be considered not valid.
3. Added setAllowValidationWhileEditing(true) for validation while editing
4. Also fixed theme change handling to update invalid qvalidated colors correctly (knots specific)
5. Fixed checkbox-dependent saving behavior for proxy settings - previously empty or invalid values were saved, even if the associated checkbox was unchecked. Now, if the associated checkbox is unchecked, values (whether valid or invalid) are ignored.
6. Added QT6 check for connecting socks and tor proxy checkboxes properly. follows the same pattern that @luke-jr used for the prune checkbox. Needed for calls to qvalidatedlineedit setEnabled() to work properly.

[1440000bytes]

Tested v29.1 and PR branch. This pull request fixes the proxy validation in GUI although I am not sure about the approach.

[1440000bytes]

Note: We can use multiple proxies in v29.1 although GUI doesn't support this yet and it would be out of scope for this pull request.

https://github.com/bitcoinknots/bitcoin/blob/v29.1.knots20250903/doc/release-notes.md#updated-settings

[bigshiny90]

i'm glad you posted the screenshots.. it pointed out an edge case that should NOT be happening - the proxy fields (QValidatedLineEdit widgets) did NOT originally validate until after focusOut. This leads to an edge case where you can input invalid IP or Port and then click OK to save the invalid data.

This is fixed in this latest version.

Also, found a bug where the checkboxes connect code was incorrect with >=QT6.7.. Luke has a fix for this already for OptionsDialog 'prune' checkbox. Had to use this pattern to cover the connectSocks and connectSocksTor checkboxes.

Lastly, the proxyvalidator used here does not mark 'error locations' and the default behavior of QValidatedLineEdit is to always set the first character position as an error, then marking the first char of the string as invalid (red, underlined, bold). Most validators don't do this - except the bitcoin address validator. So, i've changed qvalidatedlineedit to not mark ANY chars as 'errors' as it doesn't make sense. i.e. only validators that support error locations should result in the chars in the string being marked as errors. Another approach might be to extend ProxyValidator to actually support error locations, which gives visual feedback (red text) indicating problem parts of the string. seemed overly complicated for this use case, but the possibility is there.

[Ataraxia009]

LGTM

Tested, everything looks fine

Side note:
There is a bug on how signverifymessagedialogue handles changing the window to any other app on your computer and then coming back to bitcoin knots.
But this is out of scope for this PR

[1440000bytes]

ACK 2cafcb8

[bigshiny90]

Quick note: I had been thinking the original validation code handled hostnames…but when reviewing, turns out it didn’t. Hence, after changes to use inet_pton, the original code was redundant - it was just using NumericLookup to validate numeric input, but less competently than inet pton