fix: resolve path ownership problems when using sqlx_macros_unstable

[lily-mosquitoes]

When using --cfg=sqlx_macros_unstable the codepath taken uses the path variable which was taken earlier by resolve_blocking; this is solved by having resolve_blocking take a reference, as it is more convenient than cloning.

Does your PR solve an issue?

Yes (however I did not create an issue here).
I was trying to get the effect of triggering recompilation on migration changes by setting the nightly flag as described here: https://docs.rs/sqlx/latest/sqlx/macro.migrate.html#nightly-rust-cfg-flag

But when adding to .cargo/config.toml:

[build]
rustflags = ["--cfg=sqlx_macros_unstable"]

and having the sqlx::migrate! macro in the code, the following error appears while compiling sqlx-macros-core:

error[E0382]: borrow of moved value: `path`
    |
98  |     let path = path.canonicalize().map_err(|e| {
    |         ---- move occurs because `path` has type `PathBuf`, which does not implement the `Copy` trait
...
107 |     let migrations = sqlx_core::migrate::resolve_blocking(path)?
    |                                                           ---- value moved here
...
113 |         let path = path.to_str().ok_or_else(|| {
    |                                              ^^ value borrowed here after move
...
116 |                 path
    |                 ---- borrow occurs due to use in closure
    |
help: consider cloning the value if the performance cost is acceptable
    |
107 |     let migrations = sqlx_core::migrate::resolve_blocking(path.clone())?
    |                                                               ++++++++

Alternatively I could simply clone the path variable (as suggested by the compiler), however it seems like resolve_blocking does not really need to take ownership of the path variable.

I'm not too sure how this could be tested. Please let me know if the "tests/x.py" script is a good place to setup a test for this.