add: allow withdraw SOL to PDA

[brewmaster012]

closes #28

tested by anchor test

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 8.60%. Comparing base (98bd5c3) to head (7ca038b).

Additional details and impacted files

@@          Coverage Diff          @@
##            main     #37   +/-   ##
=====================================
  Coverage   8.60%   8.60%           
=====================================
  Files          1       1           
  Lines        244     244           
=====================================
  Hits          21      21           
  Misses       223     223           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[skosito]

is there a way to expand this comment a bit? account is still marked as mutable, and writing to unchecked accounts (which doesnt check owner of account) seems more dangerous than reading

maybe we can add explanation why writing to it is not dangerous, but up to you if it makes sense

[brewmaster012]

i'll keep it simple in the comment, but expand here:

Account<'info, T> is an Anchor convenient struct that automatically does "ownership" check, which UncheckedAccount explicitly does not.

Ownership is usually critical because these accounts are completely supplied by user (tx builder) and
could be anything. The onwership validation makes sure the user supplied the correct account.
For example, in the above pub pda: Account<'info, Pda>, Anchor does ownership check
Account. info. owner == T::owner(), which checks whether the supplied PDA account is in fact
owned by the current program (i.e. created by the current program). This is crucial otherwise
the program would be directed to read any user supplied potential malicious PDA account.

For the to field, obviously we don't care who its owner is because we don't read it. Whatever
tx builder tells us to send to, we happily send to. In this case Anchor requires program to explicit
exempt ownership checking by the UncheckedAccount struct and also the comment ///CHECK.

[lumtis]

Agree on adding more context in the comment as well