The update of MobileSAM and other code optimizations.

[shinning0821]

• Added the "-encoder" option, which allows specifying the specific ImageEncoder type to be used in addition to the "-net" option for specifying the network type.
• Now it is possible to use the Adapter for quick model migration or train from scratch by selecting the -mod option in the command line.
• The framework now supports MobileSAM, which can be accessed through the command line using "-net mobile_sam". It supports the use of vit, tiny_vit, and efficient_vit as encoders. Additionally, when using vit or tiny_vit, it supports the use of Adapter (adapter support of efficient_vit will be released later due to its specific structure). It also supports using input images with different resolutions. Training and testing success on ISIC and REFUGE.
• Other code refactoring and optimization.