Fix TinyMatrix determinant for N>3

See merge request !53

The case of zero determinant was not treated correctly when dealing
with N*N matrices with N>4

For these matrices, a Gauss-pivot is used but it was incorrect in the
singular case

Issue/21

Closes #21

See merge request !52

This allows to simplify the code for function data builder

Natural conversion checker can now optionally accept the case R->R^1

Fixes #21

Seems to work for functions that maps to a single space (not a list)

Related to issue #21

This avoids a few code duplication

From the ast point of view this gives cleaner data type description.
Nodes used to describe type names are now explicitly described as
type names. Previously,

- the data type could be marked as the described type. Now it is
referred as the `typename` of the type. For instance the `R^2`
type descriptor node was given `R^2` as a data type. Now it is given
the `typename(R^2)`: it is a typename which refers to `R^2`.
- In some other cases, it could have been just `typename` which forced
to re-parse entries (which is never good).

Also `affectations` types is fixed: it is now given the `void` type
instead of the type of the affected data.

Functions input and output spaces types should now be correctly
defined: some more tests are however required.

The case of compound types is however not that clear since the data
type of the type `R*R*R^2` is just defined as `typename(list)`. This
may be improved (in order to help checking of function definition/use.

An initialization issue related to tuples of R^1 was fixed on the way
and some error messages improved.

This is related to issue #21

Produce a simple configuration report at CMake invocation

See merge request !51

Also removed explicit warnings messages. This warning had a spurious
effects in IDE compilation (counted as compilation warnings for
instance)

Update/gitlab ci

See merge request !50

Coverage is now performed using gcc10/fastcov

Also test dag functionality

New requirements
- CMake 3.10
- gcc 8
- clang 8

Update/packages

See merge request !49

subrepo:
  subdir:   "packages/Catch2"
  merged:   "2b34b5c7d"
upstream:
  origin:   "git@github.com:catchorg/Catch2.git"
  branch:   "master"
  commit:   "2b34b5c7d"
git-subrepo:
  version:  "0.4.1"
  origin:   "git@github.com:ingydotnet/git-subrepo.git"
  commit:   "a04d8c2"

Add a static switch to choose verbosity of PCG/BiCGStab

See merge request !48

Feature/mass center

See merge request !47

It is replace the iso-barycenter when using MeshData::xj()

Iso-barycenter is now accessed through MeshData::cellIsoBarycenter() function

note that the functions of MeshData will be eventually renamed

Feature/free boundary condition

See merge request !46

The aim of this kind of boundary condition descriptor is to indicate
that as "special" treatment is to be performed for a boundary that
is not subject to a boundary condition at the continuous level.

For instance it may occur when using a centered scheme to discretize a
system which has only one wave (quite often with scalar laws:
transport, traffic flow,...)

It can be used both to change locally the scheme and to check that no
boundary has been forgotten in the description

Feature/diamond dual mesh manager

See merge request !45

Now the polyhedral connectivity is only stored when needed (that is
when general polyhedrons are present). This improves a lot the
creation of output files.

Feature/glace boundary conditions

See merge request !44

Now BC types are
- Dirichlet
- Fourier
- Neumann
- Symmetry

One can set tags for Dirichlet, Neumann and Fourier in order to
distinguish variables concerned by these conditions.

In the future, it is planned to allow freefem like conditions such
as `u = g on XMIN` or `alpha * u + dnu (u) = g on 3`,...

However this will only make sense when "algorithms" will be set in
pugs' language.

Now one checks that A(i,j) is accessed with i being the correct
ItemId and j being an integral type. This avoids access such as
- A(i,j) with i being an integral type
- A(i,j) with j being an ItemId

Also add bounds checking for the second index `j` (previously the only
bound checking was if (i,j) was in the scope of the data set, checked
by Kokkos). One also check that `i` is a correct ItemId value, it was
not checked in all cases.

Feature/glace boundary conditions

See merge request !43

Observe that acoustic solver implementation still requires a lot of clean-up