Not sure where trim potentiometer got into this. A variable resistor for high load would not be a trim potentimeter, but something like a rheostat.

But I'm not sure a wheatstone bridge is suitable for this. It is good for determine the value of one of the four components, where the other three are known.

In this case, the resistance is already known. The problem is to measure the current.

When switching range, you will not be able to switch resistors instantly, and with your 10k samples/second, you will have many failed measurements until you have adjusted the shunt if you use a break before make solution.

If you have a chain:

+-0E001 -+- 0E009 -+- 0E09 -+- 0E9 -+- 9E -+- 90E -+- 900E -+...
|        |         |        |       |      |       |        |
0        1         2        3       4      5       6        7

Then you can connect to the next higher or next lower connetor before disconnecting the other connector. You will get an instant switch of a factor 10 since your selector will short-circuit - or stop short-circuit the next resistor.

When processing the samples, you will be able to detect the range change as a jump of a factor 10 plus or minus a bit of turbulence from contact bounce when connecting to a lower decade or disconnecting a higher decade, and a bit of ringing or ramp when the current generator adjusts.

An alternative design is:

+-- 0E001 ---o 1
|
+-- 0E01 ----o 2
|
+-- 0E1 -----o 3
|
+-- 1E ------o 4
|
+-- 10E -----o 5
|
+-- 100E ----o 6
|
+-- 1k ------o 7
|
0

But with this design, you will get R // 10R for a while if you implement a make before break solution, i.e. for a while you will have a 10% measurement error. If using break before make, then this alternative is better than the previous, since you send the current through fewer components.