Nominal support for wide char params binding

[staticlibs]

Edit: up to date description:

This change adds only a nominal support for passing query parameters
as SQL_WVARCHAR, it uses existing logic that is already there for
SQL_WCHAR parameters.

It allows to use parameters from PyODBC, but in current impl only ASCII
data can be passed in these parameters.

Testing: none, test suite needs to be extended to test unicode through
the driver manager (related issue - #69).

Original description (encoding conversion removed in PR update):

Before this change the binding for SQL_WCHAR parameters was declared in ParameterDescriptor::SetValue but was not actually implemented (no encoding conversion).

This change implements support for getting incoming SQL_WCHAR/SQL_WVARCHAR parameters as normal UTF-8 values. Incoming UCS-2 (on Windows) or UTF-32 (on other platforms) parameter data is converted to UTF-8 using C++ stdlib.

Testing: added a test that inserts non-ASCII data as literal and then uses wide char parameter binding to match this data in SELECT.

Fixes: #49

[staticlibs]

Apaprently this does not work with Windows ODBC driver manager that uses Unicode-to-ANSI mapping with DuckDB driver. It seems that double conversion is happening, moving this PR to Draft, will follow up on it.

[staticlibs]

Updated the PR leaving only the nominal support for SQL_WVARCHAR to not throw error on PyODBC parameters binding. Filed #69 for proper unicode support.

Encoding conversion is removed, the C++ stdlibs seem to be inadequate for this task because it throws on failure. But the correct way should be to do input conversion as lenient as possible (with optional strict mode) to not break users' ETL pipelines because of invalid UTF-8 sequence in one field.

Also testing with linking directly to driver does not seem to be useful, some wider testing setup is needed that goes though the driver manager.