Fix --fmt with --by-percent producing zeros and warnings (#967)

[AnasImloul]

Summary

Fixes #967.

When --fmt is combined with --by-percent, cloc round-trips its results through an intermediate JSON. That JSON uses blank_pct / comment_pct / code_pct keys, but print_format_n only reads the canonical blank / comment / code. The result was a stream of Use of uninitialized value warnings on STDERR and a table of zeros.

Repro (before this PR)

$ cloc --fmt=4 --by-percent=cm <some dir>
Use of uninitialized value $nB in addition (+) at .../cloc line 15819.
Use of uninitialized value $nCm in addition (+) at .../cloc line 15820.
... (many more)
File      blank  comment     code    Total
./a.c         0        0        0        0
./b.py        0        0        0        0

After

File      blank  comment     code    Total
./a.c      0.00    33.33    66.67      100
./b.py     0.00    33.33    66.67      100

Changes

• cloc — in load_json:
  • Alias *_pct keys to their canonical names so the rest of the format pipeline works unchanged.
  • Fall back from a missing language field to the entry key (by-language reports don't carry one), avoiding a separate length(undef) warning.
• Unix/t/01_opts.t — regression test that asserts no Use of uninitialized value on STDERR and non-zero output for --fmt=4 --by-percent=cm.

Test plan

• Original repro from the issue produces clean output with no warnings.
• prove t/00_C.t t/01_opts.t t/02_git.t — all 952 tests pass (including 2 new).

[AlDanial]

/tmp/cloc » ./cloc --fmt 3 --by-percent=cm tests/inputs            
Illegal division by zero at ./cloc line 4892.

/tmp/cloc » ./cloc --fmt 4 --by-percent=cm tests/inputs
Illegal division by zero at ./cloc line 4892.

[AnasImloul]

@AlDanial Thanks for surfacing those failing modes. I verified that the Illegal division by zero at cloc line 4892 is pre-existing on master (i.e. not introduced by my PR, it reproduces against master with tests/inputs/issues/644/UInt8.cs, which has code=0, comment=0, blank=1, making $DEN = 0 under --by-percent=cm).

The reason it surfaces under --fmt 3/4 is that generate_report computes the txt $data_line for every output style, even when the active output is JSON, so the division at line 4892 fires before the --fmt consumer ever sees the JSON.

Since it's in the same code path as #967, I've extended this PR to fix it too:

• Added > 0 guards (matching the existing diff_report pattern at lines 3725, 4076) to the txt, yaml/json, csv/md, and SUM branches in generate_report. When the denominator is zero, the field is emitted as 0.00 rather than crashing.
• Extended the regression test in Unix/t/01_opts.t to run --fmt=3 --by-percent=cm and --fmt=4 --by-percent=cm against the full tests/inputs corpus and assert no Illegal division by zero on STDERR.

Verified clean against:

• --fmt 3/4 --by-percent=cm tests/inputs.
• --by-percent=cm tests/inputs (no --fmt), --xml, --csv, and --by-file variants.
• prove t/00_C.t t/01_opts.t t/02_git.t.

[AlDanial]

This is good progress.

Couple of minor issues: 1) the SUM line for all --fmt values and 2) the Total column for --fmt 2, 4, and 5 are incorrect.

[AlDanial]

thank you!