Fix a bug at roundoff error in block_logical_coords

[nilsvu]

This bug appeared when interpolating between two BBH domains with bulged frustums. Points at inner frustum edges were not mapped to the source domain because the frustum inverse reports logical coordinates outside [-1, 1] by roundoff error.

Proposed changes

Upgrade instructions

Code review checklist

• The code is documented and the documentation renders correctly. Run
  make doc to generate the documentation locally into BUILD_DIR/docs/html.
  Then open index.html.
• The code follows the stylistic and code quality guidelines listed in the
  code review guide.
• The PR lists upgrade instructions and is labeled bugfix or
  new feature if appropriate.

Further comments

[kidder]

please add a test that would fail/pass before/after this change

[markscheel]

Should this bug be fixed in Frustrum instead? (That is, should Frustrum be changed to never return logical coords outside of [-1,1]? ).

I worry that this change ignores #3913 and similar places where we assume logical coords are 1.0 and/or -1.0 on the boundary.

Or maybe #3913 and other places need to account for roundoff differently.

[nilsvu]

Many map inverses currently return values outside -1,1. Yea, maybe we should require all maps to ensure they return only values within their range. That's a lot of work though and needs to be tested thoroughly. Not all maps inverses return logical coords either, right? I haven't made up my mind on what's the best way to handle this.

[kidder]

• So as @markscheel says checking exact vs checking within roundoff should probably be a choice of the user of this function.
• The testing of maps needs to be done on the maps held by the Blocks, which are usually a composition of multiple maps, not individual maps.

[nilsvu]

This is now ready to review for whoever wants to @kidder @markscheel @nilsdeppe @wthrowe

[nilsdeppe]

LGTM