Regression in v0.7.5 when loading a CSV with only column names

[Octogonapus]

I have noticed a regression introduced in v0.7.5 when loading a CSV containing only column names.

Given a file foo.csv (containing only column names):

a, b

In CSV v0.7.4,

julia> import CSV
julia> using DataFrames
julia> DataFrame(CSV.File("foo.csv"))
0×2 DataFrame

In CSV v0.7.5,

julia> import CSV
julia> using DataFrames
julia> DataFrame(CSV.File("foo.csv"))
0×0 DataFrame

I am using Julia version 1.4.2. I get the same result with various versions of DataFrames v0.21.x.
CC @ianshmean

[quinnj]

This was fixed by JuliaData/DataFrames.jl#2341; I think @bkamins is planning a new DataFrames release w/ the fix soon