fix(backtest): normalize freq before querying benchmark feature data

[Pujitha Paladugu (pujitha24)]

Description

get_higher_eq_freq_feature() (in qlib/utils/resam.py) now normalizes the
freq argument (e.g. "1day" -> "day") before querying D.features, so
that a freq string equivalent to but spelled differently than the freq used
to name feature files on disk still finds the data.

Motivation and Context

Fixes #1925

When a backtest executor is configured with time_per_step="1day", that raw
string is threaded through to PortfolioMetrics/Indicator and on into
_cal_benchmark() in qlib/backtest/report.py, which calls
get_higher_eq_freq_feature(..., freq="1day"). Feature bin files on disk are
named using the normalized freq string (see
FileFeatureStorage.__init__ in qlib/data/storage/file_storage.py, which
builds file_name from the raw freq, matching files like $close.day.bin).
Since "1day" and "day" are the same frequency but different strings, the
query for "1day" silently finds no data (it doesn't raise — the freq/support
check in FileStorageMixin.uri treats them as equal via Freq.__eq__, but
the literal filename lookup does not), and the resulting empty DataFrame
trips the len(_temp_result) == 0 check, surfacing as the misleading error
ValueError: The benchmark [...] does not exist. Please provide the right benchmark — even though the benchmark data is present under day freq.

Note this is a silently-empty-query / misleading-error bug, not a crash: the
existing try/except (ValueError, KeyError) fallback in
get_higher_eq_freq_feature was never reached here because no exception was
raised in the first place.

How Has This Been Tested?

• Added a targeted regression test, tests/misc/test_resam.py, which
  mocks qlib.data.data.D and asserts get_higher_eq_freq_feature(freq="1day")
  queries with the normalized freq="day". Verified with git stash that
  this test fails on the pre-fix code (AssertionError: '1day' != 'day') and
  passes after the fix.
• cd tests && python -m pytest misc/ -v — 13 passed (includes the 2 new
  tests plus all pre-existing tests in that directory; no regressions).
• black qlib/utils/resam.py tests/misc/test_resam.py -l 120 --check --diff — clean.
• pylint (repo's configured ignore list) on qlib/utils/resam.py — 10.00/10.
• flake8 (repo's configured ignore list) on qlib/utils/resam.py — clean.
• mypy qlib/utils/resam.py — 2 pre-existing errors reported, both on lines
  untouched by this change (unrelated to the fix).
• Not run: the full CI matrix (make dev, full pytest tests/ against
  downloaded CN market data, sphinx docs, nbconvert) — this requires
  downloading several GB of market data and was impractical to run in this
  environment. The change is a small, self-contained string-normalization fix
  fully exercised by the targeted unit test above without needing live market
  data.

Screenshots of Test Results (if appropriate):

1. Pipeline test: not run (see "Not run" note above).
2. Your own tests: python -m pytest tests/misc/test_resam.py -v — 2 passed
   (fails on pre-fix code, passes after the fix).

Types of changes

• Fix bugs
• Add new feature
• Update documentation

Fixes #1925