[DRAFT] Add AOT JIT+LTO capability

[divyegala]

Description

This PR is a POC to add AOT JIT+LTO capability by using the cuVS architecture. The initial POC is on murmurhash_x86_32 functionality, which uses a device-side type dispatcher. The idea behind LTO for this functionality is that we generate a hashing device function fragment for each cudf type, and then only link the fragments for the types present in the input table. For the other types, a no-op fragment is linked instead.

Benchmarks:

## [0] NVIDIA GB10

|  num_rows  |  nulls  |     hash_name      |   Ref Time |   Ref Noise |   Cmp Time |   Cmp Noise |        Diff |   %Diff |  Status  |
|------------|---------|--------------------|------------|-------------|------------|-------------|-------------|---------|----------|
|   65536    |    0    | murmurhash3_x86_32 | 211.529 us |      91.43% | 224.138 us |      94.74% |   12.609 us |   5.96% |   SAME   |
|  16777216  |    0    | murmurhash3_x86_32 |   3.602 ms |       7.98% |   2.890 ms |       8.87% | -711.999 us | -19.77% |   FAST   |
|   65536    |   0.1   | murmurhash3_x86_32 | 457.605 us |      90.61% | 262.888 us |     120.01% | -194.717 us | -42.55% |   SAME   |
|  16777216  |   0.1   | murmurhash3_x86_32 |   3.990 ms |      11.70% |   3.243 ms |      11.95% | -746.530 us | -18.71% |   FAST   |

Checklist

• I am familiar with the Contributing Guidelines.
• New or existing tests cover these changes.
• The documentation is up to date with these changes.

[devavret]

Is this overlapping #22209 or meant as an alternative approach?

[divyegala]

@devavret it is meant as an alternative approach. This is the approach we are using in cuVS in production currently, designed by @robertmaynard, @KyleFromNVIDIA, and myself.