Promote missing_fragment_specifier to hard error

[matklad]

It has been deny_by_default since 2017 (and warned for some time
before that), so it seems reasonable to promote it.

The specific technical motivation to do this now is to remove a field
from ParseSess -- it is a global state, and global state makes
extracting libraries annoying.

Closes #40107

[rust-highfive]

r? @varkor

(rust_highfive has picked a reviewer for you, use r? to override)

[matklad]

r? @petrochenkov

[petrochenkov]

The specific technical motivation to do this now is to remove a field
from ParseSess -- it is a global state, and global state makes
extracting libraries annoying.

You remove one today, tomorrow two more are added for diagnostics.
Have to run fast just to stay in the same place.

@bors r+

[bors]

📌 Commit 91ba6d47ae593573f943dc5e5433e6d11682c046 has been approved by petrochenkov

[Mark-Simulacrum]

@bors rollup=never (could have perf implications though I don't really expect it)

[bors]

⌛ Testing commit 91ba6d47ae593573f943dc5e5433e6d11682c046 with merge 010fa54dba0d327f682636d6702119548e5fc923...

[tmandry]

@bors retry
yielding to rollup

[bors]

⌛ Testing commit 91ba6d47ae593573f943dc5e5433e6d11682c046 with merge 64295ac692af2e382fe3b749c19d554ba4e2320f...

[bors]

💔 Test failed - checks-actions

[matklad]

The failure is genuine: cargotest tests a https://github.com/dzamlo/treeify project, which hasn't been updated in four years and depends on the old clap, which has a genuine bug in the macro definition.

I think we should remove treeify from cargotest (in comparision to other projects tested, it does feel like an outlier with only a single dep (clap) and small size).

However we might want to maybe run a crater here?

[matklad]

Hm, for some reason cargotest tests fails locally for me with some git error (?!), so let's

@bors try

[bors]

⌛ Trying commit a07e52b22158978c7e6282c8c0ad183fde66af70 with merge 47ca8d06b142ea091021474a428f13a79a943058...