✨[Feature] Centralize defaults for compilation within core

[gs-olive]

Overview

Define all default values used for compile functions within a centralized location within the core library. Specifically, for the TS path which connects to the core library through Python and C++ APIs, the defaults should be located in a centralized location to avoid confusion with differing values between Python and C++ and to allow other functions within core to use these values.

Example

min_block_size will be used as an example of a compilation field for which the defaults exist in multiple locations. Specifically, for Python:

TensorRT/py/torch_tensorrt/ts/_compiler.py

Line 31 in 5fa6374

min_block_size=3,

For C++:

TensorRT/cpp/include/torch_tensorrt/torch_tensorrt.h

Line 695 in 835abf0

uint64_t min_block_size = 3;

Both of these fields ultimately populate an entry in the internal PartitioningInfo structure, which is common to both APIs. Thus, code understandability would be improved by storing such defaults in a single location. Additionally, there are times when other functions could benefit from using these defaults. For example, in partitioning, there is a specific case in PR #1599 where the default min_block_size is needed, but the value cannot be directly accessed easily, making the code less adaptable.

[github-actions]

This issue has not seen activity for 90 days, Remove stale label or comment or this will be closed in 10 days