External force extension

[LudwigBoess]

Added the option to access particle properties in the external force computation. This allows to compute more than just external force fields.

[haykh]

@LudwigBoess i think this is a bit too destructive and backwards incompatible. let's go with something like this:

define a new trait in global/traits/archetypes.h:

namespace traits::fieldsetter {
  template <class T, Dimension D>
  concept HasFx1WithIndex = requires(const T& t, const coord_t<D>& x_Ph, prtlidx_t p) { ... };
  // note that we are NOT passing particle arrays here.
}

change the ExtFieldPolicyClass to reflect this new trait:

concept ExtFieldPolicyClass = ... or ::traits::fieldsetter::HasFx1WithIndex<F, D> ...;

then in the pusher you can have:

static constexpr auto HasExtFx1 = ::traits::fieldsetter::HasFx1<F, D>;
static constexpr auto HasExtFx1WithIndex = ::traits::fieldsetter::HasFx1WithIndex<F, D>;

// ... then when calling in the `getExternalForce`:

if constexpr (HasExtFx1WithIndex) {
  f_x1 = policies.external_fields_policy.fx1(x_Ph, p);
} else if constexpr (HasExtFx1) {
  f_x1 = policies.external_fields_policy.fx1(x_Ph);
} ...

Now in the problem generator, when you define the external fields class with the new caller:

template <Dimension D>
class MyExtField {
  ParticleArrays prtls;
  
  MyExtField(const ParticleArrays& prtls) : prtls { prtls } {}

public:
  Inline auto fx1(const coord_t<D>&, prtlidx_t p) const -> real_t { ... }
};

// then in pgen class:

auto ExternalFields(simtime_t time, spidx_t sp, const Domain<S, M>& domain) const -> std::pair<bool, MyExtFields<M::Dim> {
  return { true, MyExtFields<M::Dim> { domain.species[sp - 1] } }; // this will cast Particles object to ParticleArrays which is serializable
}

And that should be it. the old functionality will still work while allowing for additional flexibility. Do you wanna give it a shot, or should i do that?