Avoid interpolating from guard cells in BTD (#5342)

BTD diagnostics sometimes show artifacts at the edge of the range of
collected data. (See for instance the red curve below.)

My understanding is that this happens because the BTD collection planes
use data from the guard cells outside of the simulation domain, when
interpolating fields that are then used for the Lorentz back-transform.
The guard cell data may not be physically correct (e.g. it may not have
the right cancellation between `E` and `B`), and could thus cause this
artifact.

This PR avoids this issue by prevents the collection planes to collect
data when it is half a cell from the edge of the simulation domain.

See the example below, taken from
#5337 (plot of the laser field,
from the BTD diagnostic)
![Figure
40](https://github.com/user-attachments/assets/e4549856-4182-4a87-aa26-2d3bc6ac8e2c)
The BTD diagnostics values are identical with this PR, except for the
problematic point appearing at the edge of the domain.

Regression/Checksum/benchmarks_json/test_2d_rigid_injection_btd.json

@@ -1,22 +1,22 @@
 {
+ "lev=0": {
+ "Bx": 3.719030475087696e-05,
+ "By": 0.004843257051761486,
+ "Bz": 5.522765606391185e-06,
+ "Ex": 1461264.5033270014,
+ "Ey": 11205.64142004876,
+ "Ez": 282020.7784731542,
+ "jx": 16437877.898892798,
+ "jy": 2492340.3149980744,
+ "jz": 215102423.57036853,
+ "rho": 0.7246235591902171
+ },
  "beam": {
- "particle_momentum_x": 2.2080215038948936e-16,
+ "particle_momentum_x": 2.2080215038948934e-16,
  "particle_momentum_y": 2.18711072170811e-16,
- "particle_momentum_z": 2.730924530737497e-15,
- "particle_position_x": 0.0260823588888081,
+ "particle_momentum_z": 2.730924530737456e-15,
+ "particle_position_x": 0.026082358888808558,
  "particle_position_y": 0.5049438607316916,
  "particle_weight": 62415.090744607645
- },
- "lev=0": {
- "Bx": 3.721807007218884e-05,
- "By": 0.004860056238272468,
- "Bz": 5.5335765596325185e-06,
- "Ex": 1466447.517373168,
- "Ey": 11214.10223280318,
- "Ez": 283216.0961218869,
- "jx": 16437877.898892513,
- "jy": 2492340.3149980404,
- "jz": 215102423.57036877,
- "rho": 0.7246235591902177
  }
 }

Source/Diagnostics/BTDiagnostics.cpp

@@ -999,12 +999,15 @@ BTDiagnostics::GetZSliceInDomainFlag (const int i_buffer, const int lev)
 {
  auto & warpx = WarpX::GetInstance();
  const amrex::RealBox& boost_domain = warpx.Geom(lev).ProbDomain();
+ const amrex::Real boost_cellsize = warpx.Geom(lev).CellSize(m_moving_window_dir);
  const amrex::Real buffer_zmin_lab = m_snapshot_domain_lab[i_buffer].lo( m_moving_window_dir );
  const amrex::Real buffer_zmax_lab = m_snapshot_domain_lab[i_buffer].hi( m_moving_window_dir );
 
+ // Exclude 0.5*boost_cellsize from the edge, to avoid that the interpolation to
+ // cell centers uses data from the guard cells.
  const bool slice_not_in_domain =
- ( m_current_z_boost[i_buffer] <= boost_domain.lo(m_moving_window_dir) ) ||
- ( m_current_z_boost[i_buffer] >= boost_domain.hi(m_moving_window_dir) ) ||
+ ( m_current_z_boost[i_buffer] <= boost_domain.lo(m_moving_window_dir) + 0.5_rt*boost_cellsize) ||
+ ( m_current_z_boost[i_buffer] >= boost_domain.hi(m_moving_window_dir) - 0.5_rt*boost_cellsize) ||
  ( m_current_z_lab[i_buffer] <= buffer_zmin_lab ) ||
  ( m_current_z_lab[i_buffer] >= buffer_zmax_lab );