fix: read xf/@xfId so a cellXfs entry resolves against its own style

[developer0hye]

What

CellFormat::set_attributes reads numFmtId, fontId, fillId, borderId and
all six apply* flags, but not xfId — even though the struct already carries a
format_id field for it and write_to writes it back out.

get_format_id() therefore always returns 0, so Stylesheet::make_style pairs
every cellXfs entry with cellStyleXfs[0] instead of the style it names:

let def_cell_format = self.cell_style_formats.get_cell_format()
    .get(*cell_format.get_format_id() as usize)   // always 0

Why it matters

get_style_by_cell_format lets that referenced style turn a whole formatting
category off:

let mut apply = true;
if def_cell_format.has_apply_number_format() { apply = *def_cell_format.get_apply_number_format(); }

When cellStyleXfs[0] carries applyNumberFormat="0" — which Excel writes for
the Normal style in many templates — apply is false for every cell in the
workbook and no number format is ever attached. Percent, currency and date cells
all read back as the raw stored number.

Measured on such a workbook (cell holding 0.25 under numFmtId="9",
xfId="6", where cellStyleXfs[6] also declares numFmtId="9"):

workbook	get_formatted_value()	resolved format
as shipped	"0.25"	None
+ explicit <numFmt numFmtId="9" formatCode="0%"/>	"0.25"	None
applyNumberFormat="0" removed from cellStyleXfs[0]	"25%"	Some("0%")

0 of its 294 cells resolved a number format. Adding the format definition changes
nothing — only the wrong default xf matters.

Reading the attribute also stops a round trip from rewriting every xfId as 0,
which currently loses the cell-style association on save.

Tests

Three unit tests in src/structs/cell_format.rs: a declared xfId is read from a
self-closing element and from one with children, and an absent one stays 0.

cargo test on this branch: 109 + 113 + 4 + 79 pass.

cargo clippy -- -D warnings and cargo test are green on 1.88.0, stable and
nightly; cargo +nightly fmt --all --check is clean.

Note

The second commit is unrelated to the xfId fix and is here only to get CI
green. Current nightly clippy denies nonstandard_macro_braces, so
cargo clippy -- -D warnings fails on master at src/writer/csv.rs:47:

error: use of irregular braces for `format!` macro
  --> src/writer/csv.rs:47:25

Because that job aborts the matrix, the other three toolchains are cancelled
before they finish and every open pull request reads as red. The commit
switches that one call to parentheses, which is what clippy suggests; rustfmt
then wraps the arguments, since the brace form had exempted it. Happy to split
it out if you would rather take it separately.

The third commit swaps the deprecated get_format_id / get_number_format_id
in the new tests for format_id / number_format_id, so the suite compiles
without deprecated warnings.

[MathNya]

@developer0hye
Thank you for the PR.
I'll go ahead and merge it.