Add a method to extract system message based on adapter and signature

[chenmoneygithub]

In some scenarios users want to extract the actual prompt sent to LM programmatically. Right now we allow users to use Adapter.format() to get the multi-turn message, but sometimes users only want to extract the system message. This PR introduces a new method Adapter.format_system_message() which only extracts system message based on adapter and signature.

Example usage:

import dspy

signature = dspy.Signature("question -> answer")
system_message = dspy.ChatAdapter().format_system_message(signature)
print(system_message)

The output should resemble:

Your input fields are:
1. `question` (str):
Your output fields are:
1. `answer` (str):
All interactions will be structured in the following way, with the appropriate values filled in.

[[ ## question ## ]]
{question}
[[ ## answer ## ]]
{answer}
[[ ## completed ## ]]
In adhering to this structure, your objective is: 
        Given the fields `question`, produce the fields `answer`.

[TomeHirata]

LGTM!

[TomeHirata]

Can we tag related issues if any?

[chenmoneygithub]

@TomeHirata The related issue is our work on KD service/SDK - this PR just makes it easier to extract out the prompt.