Inhibit modification hooks in ws-butler-after-save

[hlissner]

This PR sets out to fix https://github.com/seagle0128/doom-modeline/issues/129.

What this PR does: it inhibits modification hooks while ws-butler-after-save restores virtual whitespace to the buffer post-save.

Rationale:

• The doom-modeline package uses a hook in after-save-hook and after-change-functions to update a buffer-modified indicator in the mode line.
• The insert and forced move-to-column calls in ws-butler-after-save trigger after-change-functions when writing virtual whitespace into the buffer (causing doom-modeline to indicate the buffer has changed).
• There is no mechanism through which doom-modeline can subscribe to changes in the buffer's modified status besides after-save-hook, so the subsequent call to (set-buffer-modified-p nil) will not reach doom-modeline.
• Since ws-butler-after-save is appended to after-save-hook buffer-locally, the only way for doom-modeline to sneak a update hook into it by making users aware of the race condition, or accommodating ws-butler specifically.

Thus, doom-modeline incorrectly reports a modified buffer when it isn't modified.

It's debatable whether doom-modeline or ws-butler should adapt, but I didn't think insertion of virtual whitespace should trigger modfiication hooks in the first place, so here I am!

[lewang]

thanks!

I'm a bit concerned about

undo data may become corrupted

from docs of with-silent-modifications, but let's give this a whirl regardless.

[dcolascione]

I don't think this change is right. Modes are allowed to do arbitrary reparsing in response to buffer change hooks: for example, a mode could store a parse tree keyed on character counts from the start of the buffer. Modifying the buffer using with-silent-modifications can break mode parsing. doom-modeline should change instead.

[lewang]

@dcolascione Sounds reasonable. Thanks!