Specialize sparse and low cardinality behaviour for `equals` and `notEquals`

### Company or project name

CH

### Describe the situation

When processing data in filters with a sparse column, converting to and from full columns is quite expensive. This is from a query executing `!=` over a sparse column:

```
- 11.33% DB::FilterTransform::doTransform(DB::Chunk&) ▒
- 8.62% DB::ExpressionActions::execute(DB::Block&, unsigned long&, bool, bool) const ▒
- 4.99% DB::IExecutableFunction::execute(std::__1::vector > const&, std::__1::shared_ptr const&, unsigned long, bool) const ▒
- 2.67% DB::IExecutableFunction::executeWithoutSparseColumns(std::__1::vector > const&, std::__1::shared_ptr const&, unsigned long, bool) const ▒
- 1.48% DB::IExecutableFunction::executeWithoutLowCardinalityColumns(std::__1::vector > const&, std::__1::shared_ptr const&, unsigned long, bool) const ▒
- 1.22% DB::FunctionToExecutableFunctionAdaptor::executeImpl(std::__1::vector > const&, std::__1::shared_ptr const&, unsigned long) const ▒
- 1.09% DB::FunctionComparison::executeImpl(std::__1::vector > const&, std::__1::shared_ptr const&, unsigned long) const ▒
0.51% COW::immutable_ptr DB::FunctionComparison::executeNumLeftType(DB::IColumn const*, DB::IColumn const*) const ▒
0.54% DB::(anonymous namespace)::convertLowCardinalityColumnsToFull(std::__1::vector >&) ▒
0.64% DB::IColumnHelper, DB::ColumnFixedSizeHelper>::getNumberOfDefaultRows() const ▒
0.72% std::__1::vector >::__destroy_vector::operator()[abi:ne190107]() ▒
0.62% DB::Block::insert(DB::ColumnWithTypeAndName) ▒
0.53% DB::ColumnSparse::~ColumnSparse() ▒
```

This is the default behaviour for most functions because it doesn't make sense to specialize each function to work with sparse / low cardinality columns, as the complexity of the code would be too much.

On the other hand, there are some functions where making use of the dictionaries directly would make sense, such as `equals` / `=` or `notEquals`.

### Which ClickHouse versions are affected?

All

### How to reproduce

Clickbench Q1

### Expected performance

Hopefully it should be faster. If not, don't merge the change.

### Additional context

_No response_