fix(mfusg): fix precision loss in CLN/GNC packages writing coordinates

MFUSG CLN/GNC packages were losing significant precision when writing
node coordinates, causing coordinate errors up to 3.5 km and affecting
model results.

## Root Cause
The fmt_string() method in mfusgcln.py and mfusggnc.py used %10.2e
format (only 3 significant figures) instead of the standard Util2d
format %15.6G (9 significant figures).

## Impact on Real Data
Complete line from CLN package showing precision loss:

Before (bug - %10.2e):
1  1  0  5.75e+01  1.15e+03  0.00e+00  0  0  4.01e+05  6.90e+06  1.21e+03  4.01e+05  6.90e+06  1.15e+03

After (fixed - %16.9G):
1  1  0  57.4650002  1150.19897  0  0  0  400569  6903567  1207.66394  400569  6903567  1150.19897

The precision loss caused measurable differences in model results.

## Solution
Changed fmt_string() to return "%16.9G" format, which:
- Provides 9 significant figures (exceeds float32's 7-digit precision)
- Uses adaptive format (decimal when readable, scientific when needed)
- Maintains proper spacing with 16-character width
- Consistent with other MFUSG packages (WEL/DRN/RIV use %G)

## Testing
Verified with Valle Copiapo model (142 CLN nodes, 181 stress periods).
Model results now match original with negligible differences.

Author: reneangermeyer

flopy/mfusg/mfusg.py

@@ -550,7 +550,7 @@ def fmt_string(array):
         if vtype in {"i", "b"}:
             fmts.append("%10d")
         elif vtype == "f":
-            fmts.append("%10.2e")
+            fmts.append("%16.9G")
         elif vtype == "o":
             fmts.append("%10s")
         elif vtype == "s":