[Dev] PandasAnalyzer doesn't actually need pandas (duckdb#49)

This PR fixes duckdb#48

It just sniffs objects to determine if they adhere to a structured type,
that doesn't necessarily need pandas
This is kind of untested because all our tests require pandas+numpy

But I went through the logic called by the pandas analyzer and it seems
like pandas/numpy are loaded anyways when we first try to compare with a
pandas/numpy class type (i.e there's no `import_cache.pandas.NaT(false)`
in the path)

Author: evertlammerts

src/duckdb_py/pandas/analyzer.cpp

@@ -502,12 +502,6 @@ bool PandasAnalyzer::Analyze(py::object column) {
 	if (sample_size == 0) {
 		return false;
 	}
-	auto &import_cache = *DuckDBPyConnection::ImportCache();
-	auto pandas = import_cache.pandas();
-	if (!pandas) {
-		//! Pandas is not installed, no need to analyze
-		return false;
-	}
 
 	bool can_convert = true;
 	idx_t increment = GetSampleIncrement(py::len(column));