Improve lim -t assignment functionality.

Requesting enhancements to how LSF Parses and assigns host types for both x86 and POWER architectures with the lim -t command. Current behavior causes inconsistent cpuf values due to bogomips changes on x86 and insufficient CPU-model differentiation on Power9 systems. Details following on this 2-part problem with lim -t command. 

1. x86 Host-Type Instability (Due to Bogomips Changes)

Details: On x86 systems, the bogomips value reported at boot is not consistent. Most hosts present AT LEAST *two possible* bogomips values. When the value changes, LSF assigns an incorrect host type because the corresponding `lim -t` entry in `lsf.shared` no longer matches. This results in `cpuf` being misconfigured or assigned to a similar—but still incorrect—host type. We've noticed that the lim -t value assignment is partially based on the bogomips value. This number can change; therefore, if the host is rebooted and the "!" is used in the cpuf column, the lim -t assignment reconfigures to a different value, often leading to inaccuracies.

Current suggested work around: Manually listing *all possible bogomips variants* for each machine in `lsf.shared`.

Impact: 
*   Host classification becomes inconsistent after reboots
*   `cpuf` values are frequently incorrect
*   Manual admin effort increases significantly in heterogeneous clusters

Real world example: 

This is how we understand how lim -t is assigned
Example Host Architecture assigned lim -t: x6_4607_IntelRXeonRGold6140CPU230GHz 
"x6" =x86 identifier
"4607"= bogomips value (has changed int the past)
"IntelRXeonRGold6140CPU230GHz " = proc/cpuinfo value

2. POWER Systems Not Differentiated by Model Variant

Details: We operate multiple POWER9 variants (e.g., POWER9\_22A, POWER9\_42A). LSF currently parses the architecture string as a single type (e.g., `POWER9architectedaltivecsupported`) and does not distinguish between POWER9 sub-models. LSF supports Power systems. As more Power boxes were released (specifically Power 8, 9, and 10), Power is a bit special in that it isn't just a power box. Naming convention (22A and 42A) among other things can be diversly different among the machine it is still a power # box. Unfortunately, lim -t does not differentiate between Power 9_22A and Power 9_42A, resulting in machines being assigned a lower cpuf value. This misassignment means that high-priority tasks requiring better CPU resources cannot target our more capable machines.

Current suggested work around: Manually hard-coding CPU model definitions in `lsb.cluster`.

Impact:
*   Loss of precision in system reporting
*   Limitations on users' ability to properly target machines. 
*   POWER9 machines with different characteristics are treated as identical
*   Inconsistent resource matching and scheduling accuracy

Real world example: 
For POWER This is how we understand how lim -t is assigned
Example Host Architecture assigned lim -t : POWER9architectedaltivecsupported 
"POWER9architectedaltivecsupported "= proc/cpuinfo value
 

 Last Note: Overall, without better functionality of lim -t it defeats the purposed of the lim -t command auto assigning with the "!" in the lsb.cluster files. We have a lot of different machines and having an accurate lim -t output would make things a lot easier in our diverse cluster for administration and for users.