when attempting to load a lockfile and finding only a manifest, automatically generate a lockfile and load that

[toph-allen]

This pull request changes the behavior of renv_lockfile_load(), which attempts to load a lockfile from a directory. When no lockfile is found, but a manifest.json is found, the manifest is used to automatically generate a lockfile, which is then loaded. The behavior only occurs when the function is called with strict = FALSE, the default. If no manifest is present, the function renv_lockfile_init() is called, as before. A message is printed when this happens.

I added a supporting function in paths.R to return a manifest path, just cribbing from renv_paths_lockfile(). However, just doing file.path(project, "manifest.json") might be just as good. I'm not sure what the semantics are around path loading, so was just trying to follow existing patterns.

The logic in renv_manifest_path() is similar to renv_paths_lockfile(), or at least, the default case. I'm unsure about needing to add an extra layer of abstraction similar to renv_lockfile_path() -> renv_paths_lockfile().

When testing, I saw a duplicated log line:

No lockfile found; creating from `manifest.json`.
- Lockfile written to "/private/var/folders/q7/1b7h_nx931q6mc70r3w8k_2c0000gp/T/RtmpsmRCq6/file62e9cace1/renv.lock".
- Lockfile written to "/private/var/folders/q7/1b7h_nx931q6mc70r3w8k_2c0000gp/T/RtmpsmRCq6/file62e9cace1/renv.lock".

This came from explicit logging in renv_lockfile_from_manifest(), which duplicated a log message within renv_lockfile_write(), so I removed the duplicated outer logging statement.

[kevinushey]

Overall LGTM; main nits are around style + suggesting we don't need the extra function for finding the path to manifest.json

[kevinushey]

LGTM!

[toph-allen]

@kevinushey I forgot to update the NEWS.md — would you like me to make another PR with that change or are you fine handling it?

[kevinushey]

I'll take care of it, thanks!