FEAT: Add High performance Field Translation and FMM crates. (#126)

* Add field translation traits and implementations (both optimised SVD and FFT field translations are added)
* RLST is used for all underlying data structures.
* Single node kiFMM implementation, and associated traits for FMMs
* More tests added to ensure correctness of trees, and correct ordering for points created in surfaces.
* Documentation added for Field and Fmm crates, documentation updated in Tree crate where things have changed.
* Benchmark for FFT FMM (order 6, Laplace, 1e6 pts) ~1.5s (Apple M1).

.gitignore

@@ -18,3 +18,5 @@ examples.sh
 
 _*.msh
 *.btr
+
+.vscode/

Cargo.toml

@@ -4,6 +4,7 @@ members = [
     "bem",
     "element",
     "fmm",
+    "field",
     "grid",
     "hyksort",
     "quadrature",
@@ -13,10 +14,13 @@ members = [
     "traits",
 ]
 
+resolver = "2"
+
 default-members = [
     "bem",
     "element",
     "fmm",
+    "field",
     "grid",
     "hyksort",
     "quadrature",

bem/src/assembly/batched.rs

@@ -1,6 +1,4 @@
 use crate::function_space::SerialFunctionSpace;
-use bempp_kernel::traits::Kernel;
-use bempp_kernel::types::EvalType;
 use bempp_quadrature::duffy::quadrilateral::quadrilateral_duffy;
 use bempp_quadrature::duffy::triangle::triangle_duffy;
 use bempp_quadrature::simplex_rules::simplex_rule;
@@ -11,6 +9,8 @@ use bempp_traits::bem::{DofMap, FunctionSpace};
 use bempp_traits::cell::ReferenceCellType;
 use bempp_traits::element::FiniteElement;
 use bempp_traits::grid::{Geometry, Grid, Topology};
+use bempp_traits::kernel::Kernel;
+use bempp_traits::types::EvalType;
 use bempp_traits::types::Scalar;
 use rayon::prelude::*;
 use rlst_dense::{RandomAccessByRef, RawAccess, RawAccessMut, Shape};