Crystal matrix representations are shown too imprecisely

[dagewa]

The str representation of a Crystal shows matrix elements with "% 5.4f" format, like this:

>>> print(Crystal((0.02,0,0,0,0.02,0,0,0,0.02), "P1"))
Crystal:
    Unit cell: 50.000, 50.000, 50.000, 90.000, 90.000, 90.000
    Space group: P 1
    U matrix:  {{ 1.0000,  0.0000,  0.0000},
                { 0.0000,  1.0000,  0.0000},
                { 0.0000,  0.0000,  1.0000}}
    B matrix:  {{ 0.0200,  0.0000,  0.0000},
                {-0.0000,  0.0200,  0.0000},
                {-0.0000, -0.0000,  0.0200}}
    A = UB:    {{ 0.0200,  0.0000,  0.0000},
                { 0.0000,  0.0200,  0.0000},
                { 0.0000, -0.0000,  0.0200}}

If we construct a Crystal that differs from this by the least significant digit of the displayed A matrix values we get a rather notable change in the unit cell:

>>> print(Crystal((0.0201,0,0,0,0.0201,0,0,0,0.0201), "P1"))
Crystal:
    Unit cell: 49.751, 49.751, 49.751, 90.000, 90.000, 90.000
    Space group: P 1
    U matrix:  {{ 1.0000,  0.0000,  0.0000},
                { 0.0000,  1.0000,  0.0000},
                { 0.0000,  0.0000,  1.0000}}
    B matrix:  {{ 0.0201,  0.0000,  0.0000},
                {-0.0000,  0.0201,  0.0000},
                {-0.0000, -0.0000,  0.0201}}
    A = UB:    {{ 0.0201,  0.0000,  0.0000},
                { 0.0000,  0.0201,  0.0000},
                { 0.0000,  0.0000,  0.0201}}

Other tests show that a difference of 1 in the least significant digit of the "% 5.4f" representation of a rotation matrix appears to equate to approximately a 0.01° error, which isn't so bad. But imprecision in the B matrix is more problematic.

[dagewa]

This matters because with #819 the user can override a crystal model (using a forthcomingdials.modify_experiments PR) by supplying A matrix elements, but this model could be quite different from what they expect if they got the elements from dials.show output.

[dagewa]

A single digit increase in precision gets the unit cell error down to a 40th of an Angstrom, which is probably acceptable. Comments welcome.

[dagewa]

Adding a couple of digits is even better and still looks okay to me.

Crystal:
    Unit cell: 5.4753(4), 8.2107(6), 12.1584(9), 89.896(3), 90.048(3), 90.002(3)
    Space group: P 1
    U matrix:  {{ 0.699191,  0.237955, -0.674173},
                { 0.572326, -0.751420,  0.328345},
                {-0.428456, -0.615423, -0.661574}}
    B matrix:  {{ 0.182638,  0.000000,  0.000000},
                { 0.000007,  0.121793,  0.000000},
                { 0.000154, -0.000222,  0.082248}}
    A = UB:    {{ 0.127597,  0.029131, -0.055449},
                { 0.104574, -0.091591,  0.027006},
                {-0.078358, -0.074807, -0.054413}}