[compiler] Triton backend support

[6somehow]

Implement a new codegen backend triton template (abbreviated to 'tit'):

• use the PTX generated by the external TritonTemplate module
• modify the compiler module to interact with TritonTemplate
• add the support of SharedMemorySize in @PTXOp in the runtime module

[CLAassistant]

All committers have signed the CLA.

[qingyunqu]

LGTM, Thanks!