Different initial bounds (LB/UB) across different machines

Hi,

I am running NativeSARSOP using a fixed RockSample problem instance, but I am seeing vastly different initial bounds across different machines for the same script with same julia version. Not sure if this is the expected behavior.

Environment Details:

```
Julia Version: 1.12.6 (Same on both machines)
Machine A : Ubuntu 20.04.6, 11th Gen Intel(R) Core(TM) i7-11800H @ 2.3GHz
Machine B : Ubuntu 18.04.2, Intel(R) Xeon(R) Silver 4110 CPU @ 2.10GHz
```
I attempted to isolate the environment ( a temporary JULIA_DEPOT_PATH and package environment). Below are some outputs I got.

Machine A
```
--------------------------------------------------------------------------------------
Time Iter LB UB Precision # Alphas # Beliefs
--------------------------------------------------------------------------------------
0.06 0 12.5291637 27.1362835 14.6071198033 34 35
1.15 10 19.0346704 26.6821086 7.6474381954 268 306
2.94 20 19.2296742 26.0662417 6.8365674649 469 529
6.13 30 20.2018613 25.9220426 5.7201812934 667 745
11.57 40 20.5820094 25.7794694 5.1974600001 849 977
18.12 50 20.5820094 25.7468860 5.1648766479 1030 1205
27.43 60 20.5820094 25.5565391 4.9745297608 1209 1425
37.98 70 20.6102863 25.5172762 4.9069899170 1346 1625
50.17 80 20.9038401 25.4814042 4.5775641358 1496 1835
65.34 90 20.9038401 25.4290280 4.5251879738 1674 2067
81.16 100 20.9038401 25.4127152 4.5088751067 1834 2278
103.61 110 20.9038401 25.3851464 4.4813063686 1960 2479
123.00 120 21.2668382 25.3686680 4.1018298370 2105 2672
149.53 130 21.2668382 25.3332254 4.0663872047 2250 2892
177.06 140 21.2668382 25.2950156 4.0281774182 2386 3105
199.78 150 21.2668382 25.2554930 3.9886548071 2532 3330
233.40 160 21.2668382 25.2400892 3.9732510118 2687 3541
261.50 170 21.2668382 25.1689124 3.9020741724 2799 3720
307.60 180 21.3526643 25.1597583 3.8070939716 2939 3906
343.31 190 21.3526643 25.1458647 3.7932003551 3079 4112
393.34 200 21.3982445 25.1304119 3.7321673589 3240 4328
--------------------------------------------------------------------------------------
393.34 201 21.3982445 25.1304119 3.7321673589 3240 4328
--------------------------------------------------------------------------------------
```

Machine B
```
--------------------------------------------------------------------------------------
Time Iter LB UB Precision # Alphas # Beliefs
--------------------------------------------------------------------------------------
0.09 0 14.6401438 24.0443096 9.4041658681 25 29
2.50 10 18.0646099 23.0449977 4.9803878650 262 313
6.29 20 19.7505218 22.9437076 3.1931857628 458 530
12.20 30 19.7505218 22.8420942 3.0915723450 605 717
20.43 40 20.0016999 22.8001330 2.7984330881 737 921
31.13 50 20.8455431 22.7131205 1.8675773640 884 1119
43.90 60 20.8455431 22.6924725 1.8469293806 1032 1315
56.94 70 20.8455431 22.6189909 1.7734478098 1148 1479
73.15 80 20.8455431 22.5999076 1.7543644626 1240 1621
91.33 90 20.8455431 22.5580464 1.7125033115 1330 1764
111.44 100 20.8455431 22.5467181 1.7011749721 1448 1942
133.52 110 20.8704361 22.5322885 1.6618523279 1529 2102
158.26 120 20.8704361 22.4999401 1.6295039788 1614 2247
184.85 130 20.8798086 22.4837583 1.6039496844 1689 2396
214.18 140 20.8798086 22.4692229 1.5894142789 1762 2536
236.18 150 20.8798086 22.4611439 1.5813352873 1814 2660
267.76 160 20.8798086 22.4487550 1.5689464180 1878 2845
304.09 170 20.8798086 22.4456060 1.5657974625 1967 3014
338.06 180 20.8798086 22.4315795 1.5517708808 2044 3179
367.61 190 20.8798086 22.4081093 1.5283007025 2095 3314
409.19 200 20.9282403 22.3943919 1.4661515847 2172 3445
--------------------------------------------------------------------------------------
409.19 201 20.9282403 22.3943919 1.4661515847 2172 3445
--------------------------------------------------------------------------------------
```

For a fixed 60 second planning time, initially I got below output in Machine B.
```
--------------------------------------------------------------------------------------
Time Iter LB UB Precision # Alphas # Beliefs
--------------------------------------------------------------------------------------
0.26 0 12.9402387 14.2699739 1.3297352074 31 20
2.47 10 13.6936818 14.2699739 0.5762921059 98 64
--------------------------------------------------------------------------------------
2.62 12 14.9668810 14.2699739 -0.6969071009 107 1
--------------------------------------------------------------------------------------
```
The julia script is attache here [fresh_env_native_sarsop.zip](https://github.com/user-attachments/files/26998792/fresh_env_native_sarsop.zip)

Start by reproducing the attached fresh_env_native_sarsop.zip script with the fixed RockSample instance on both reported Julia and Ubuntu environments. Compare the NativeSARSOP initialization and bound calculations to determine whether platform-dependent behavior explains the different LB/UB values; done means identifying the cause and establishing expected or reproducible results.

Written by the indexing model from the issue text.